Blog

‘Web3 can fix most, if not all of the music industry’s issues when it comes to finances.’ – Music Business Worldwide

MBW’s World’s Greatest Producers series sees us interview – and celebrate – some of the greatest talents working in studios across the decades. Here we talk to !llmind, the Grammy-winning collaborator of artists including Jay Z, Beyonce, Kanye West, J Cole and Drake. !llmind also produced Lin Manuel Miranda’s Hamilton mixtape and contributed to the soundtrack for Disney’s 2016 film Moana. World’s Greatest Producers is supported by Hipgnosis Song Management.
“Everything [in] life is a gamble,” raps Kendrick Lamar on 2017’s GOD.
Those five words, taken from 2017’s DAMN., may not be the most repeated or regarded lyrics in the rapper’s repertoire.
But they’re quite fitting for the life and career of producer Ramon ‘!llmind’ Ibanga Jr.

Around eight years ago, an !llmind co-produced beat, End of the World, appeared on SoundCloud.
It was taken from !llmind’s #BoomTrap Vol. 2 mixtape released on the producer’s own Blap-Kits platform – a sample library for producers launched in 2011.
One of GOD.’s co-producers, Ricci Riera, then took a snippet of End of the World, added some drums and made one of DAMN.’s standout tracks.

Because End of the World began its life as a royalty-free sample pack on Blap-Kits, !llmind missed out on an album credit. As is the way of the producer.
!llmind is more than content with that.
He’s more focused on “bringing people together, becoming a mentor to certain people and certain communities”, he says. And besides, his other successes means that missing out on a Kendrick feature isn’t, well, the end of the world.

 
!llmind is a twice Grammy-winning, ten time-nominated music producer.
His credits include Jay Z, Beyonce, Kanye West, J Cole and Drake, he produced Lin Manuel Miranda’s Hamilton mixtape and contributed to the soundtrack for Disney’s 2016 film Moana.
With that, ordinarily, comes ego.
But !llmind is very much a producer who remembers where he came from; he’s still very much the kid who, he says, made beats in his “mom’s basement eating chocolate chip cookies”.
Today, his bedroom might be bigger, but his success remains entrenched in that mindset.
In 2021, Blap-Kits crossed the $1 million-per-year revenue mark, the company says.
Meanwhile, !llmind regularly dissects beat-making on Twitch and chews the fat on YouTube, two platforms that allow his self-proclaimed need to “nerd out about producer stuff.”
And, much like fellow producers Timbaland, 3LAU, and others, he’s entered the NFT game with his Squad of Knights project.
Squad of Knights is an NFT-powered music production community that integrates studio space, mentorship, and collaboration in the real world.
Below, you’ll find an edited/abridged conversation between MBW and !llmind that covers, amongst other things, the potential and pitfalls of NFTs, what he’s learned from producers having a bad day – and his career as one of the World’s Greatest Producers…
I feel like I change and develop every year. Back in 2005-2006 when I really started doing this, I was the kid in my mom’s basement trying to sound like J Dilla and chopping up soul records.
Now, I’ve sort of stepped into a mentorship role, but I am still very much a music producer. That’s what I am and what I will always be.
“You never know what kind of song you’ll make or if something will become a hit, but it’s that thrill of not knowing that drives me.”
When you’re a music producer, one of the biggest thrills is when you create a beat for the first time. It’s so satisfying, and even though I’ve produced 10-15 beats a day for almost 20 years, I still get that same satisfaction.
You never know what kind of song you’ll make or if something will become a hit, but it’s that thrill of not knowing that drives me. It keeps me really excited for tomorrow.
Twitch is the best at online fan engagement and providing the tools to keep people engaged. On Twitch, there’s been a recent exponential growth in the music community, and music producers specifically, because it’s just the perfect place to do it.
We’re sitting in front of our computers [for our jobs] just like gamers. Making music is essentially a gaming tool – whether it’s Fruity Loops or Pro Tools – but the difference is we’re creating music that can be monetized.
Twitch has helped me connect and open up my fan base, and it’s [enabled] me to collaborate with people. I’ll have some of my viewers submit melody loops to me, and I’ll choose one at random and we’ll collaborate right there on the spot..
There are definitely challenges sprinkled throughout my journey so far. I won’t say a name for this, but there was a project I was heavily involved with where I just couldn’t connect with the artist.
I was there making beats and collaborating, doing as much as I could, but I ended up not creating anything for the project after a couple of weeks straight. It was a huge opportunity, but I just couldn’t deliver.
“I have no expectations going into certain projects. Everything that we create is such a crapshoot.”
Another time, a while ago, I was in a room full of producers when a really big producer walked in.
I really respected and looked up to this person so much, but I just had a bad experience with them. They were really mean to everyone, and he basically kicked everyone in the room out. I don’t know if they were in a bad mood, but that session broke my heart.
Sometimes stuff like that happens, and I’m kind of numb to it. By that I mean, I have no expectations going into certain projects. Everything that we create is such a crapshoot.
You can have one track that you’re not a huge fan of that then becomes your biggest record. Meanwhile, you’re working on a record for two weeks straight, and you feel like it’s ‘the one’, and it ends up sitting on a hard drive.
You really can’t be jaded by things [in the music industry]; I’ve learned that over the years.
We need to open up communication. I don’t want to use words like ‘movement’ or producer union, but maybe we do need one.
For writers, music producers and musicians, the percentages we get, it’s pretty clear that we all see it’s unfair. I don’t know the answer to that, to be honest.
When I look at Web3, maybe that’s the solution. We’re approaching a time where Web3 technology and blockchain is a solution, technologically, for storing information contracts, getting people paid, getting people splits and distributing funds in the snap of a finger.
Web3 can do those things for you. Then there’s Web2 there like, ‘Hey, we’ve been doing it this way for 60 years, and a lot of people are making a lot of money, but you guys aren’t.’
How do you get those two to communicate? Is there a conversation that needs to happen? And who’s going to curate those conversations? We have to figure out a way for these two things to speak and coexist.
It’s a gift and a curse. All of the traffic is on Spotify, Apple and the big DSPs. They’re creating a service, but they’ve decided to say that this service is going to come at a cost.
If it were my choice, I would prefer not to go to war with them.
I would prefer a world where Web3, Web2 and DSPs could work together to solve this issue. That may be a little bit ‘Kumbaya’ for a lot of people, but fuck it.
Do you want to go to war with people who have billions of liquidity and billions of dollars and can hire the best lawyers? They will eradicate you. They’re not going to just sit back, see Web3 take over and go away.
We need to be realistic and approach this in a way where we ask, how can we fix this? We have the power, but they have a level of power too. And they have money. So we need to figure something out.
It’s a good start. I love what 3LAU is doing with Royal. [But] I don’t think that there is just one answer. We’re in a phase of exploration and trying things.
My personal opinion of it is that it’s awesome that fans get a piece of the royalty, but there’s also some issues with that. Is that adoptable by the masses? I don’t think so.
I don’t think that there are enough music fans who really give a shit about paying extra money to get royalty. A lot of fans just want to hear the music with one click on their phone, and they want to buy a ticket with one click.
It’s awesome that we’re exploring different ways for those fan interactions to happen, but I don’t think we’ve found one thing yet that’s going to solve all of our problems. But I think we’re headed in the right direction.
Squad of Knights, my NFT project, sorts out the issue of, how do I find people to work with? How do I find my community?
If I’m a rapper, how do I find beats? If I’m a music producer, how do I find my Drake? The way those success stories happen is by people meeting people, so how do we meet other people?
We have finished our public mint, and the community is growing every day.
We have a reward token called the Knights token; any time our community members collaborate, or win challenges, we reward them with Knight tokens.
We’ve partnered with some music studios in Los Angeles as a starting point where our community members can use Knight tokens to book actual studio time, meet each other, and make music together in a room.
Keep going and don’t overthink. And put all your money into bitcoin in 2011!
In 2013, I started selling sound-packs on my Shopify page.
At the time, they had rolled out an app where you could accept Bitcoin as payment. I activated it, and we made around $1,000 in Bitcoin. But there was an option where I could convert it and pay it directly into my account, or keep it in Bitcoin.
I put it into my bank account. If I had kept that wallet, I would have had well over $1,000 in Bitcoin in 2013. How much would that have been today? Maybe a couple of million.
I truly believe that the business of music would benefit if it utilized blockchain and Web3 technology.
All of these methods of collecting royalties and payments, distributing money and having long contracts, it can all be solved.
This feels like when people were using a horse and buggy, and then the car was invented. I think it’s a pretty obvious transition.
Web3 can fix most, if not all of the music industry’s issues when it comes to finances and business, but we’re yet to see when and how that transition happens.Music Business Worldwide
The best of MBW, plus the most important music biz stories on the web. Delivered for FREE, direct to your inbox each day.

source

Opinion | Russian Soldiers See That Ukrainians Live Better Than They Do – The New York Times

Advertisement
Supported by
Guest Essay
Send any friend a story
As a subscriber, you have 10 gift articles to give each month. Anyone can read what you share.

Mr. Firsov is a medic in the Ukrainian military in eastern Ukraine. He was a member of the Ukrainian Parliament from 2014 to 2016.
In early April I walked into Andriivka, a village about 40 miles from Kyiv, with my battalion in the Ukrainian territorial defense forces. We were among the first Ukrainian troops to enter the village after a Russian occupation that had lasted about a month. Shell casings and boxes of ammunition were scattered everywhere, and the houses were in various states of ruin. In one of the yards we passed there was an abandoned burned-out tank sitting on the grass.
The Russians killed civilians in Andriivka, and they ransacked and looted houses. The locals told us something else the Russians had done: One day they took mopeds and bicycles out of some of the yards and rode around on them in the street like children, filming one another with their phones and laughing with delight, as if they’d gotten some long-awaited birthday present.
A few days earlier we were in Bucha, a suburb northwest of Kyiv that was subjected to an infamously brutal occupation. The people there told us that when the first Russian convoy entered the town, the troops asked if they were in Kyiv; they could not believe that such idyllic parks and cottages could exist outside a capital. Then they looted the local houses thoroughly. They took money, cheap electronics, alcohol, clothes and watches. But, the locals said, they seemed perplexed by the robotic vacuum cleaners, and they always left those.
One resident, who told me that she was taken hostage by the Russian soldiers in her house, said they could not get over the fact that she had two bathrooms and kept insisting that she must have more people living with her.
This war is Vladimir Putin’s fatal mistake. Not because of economic sanctions and not because of the huge losses of troops and tanks but because Mr. Putin’s soldiers are from some of the poorest and most rural regions of Russia. Before this war, these men were encouraged to believe that Ukrainians lived in poverty and were culturally, economically and politically inferior.
Now the invaders have seen the reality: The Ukrainians live better than they do.
The war has brought that reality home to me, too. When Ukraine and Russia left the Soviet Union about 30 years ago, we had the same resource-based economies, the same endemic corruption and poverty. Foreigners would often think Ukraine was part of Russia, and even Ukrainians did not always understand the fundamental differences between Russians and us. But at some point, our paths diverged.
I grew up in Avdiivka, in Donetsk Oblast in eastern Ukraine. When I fell ill as a child and had a high temperature, my grandmother called an ambulance. When it didn’t arrive, we called again, and they told us they did not have enough gas to reach us. I remember that hospital, with its floor covered in tattered and bristling Soviet linoleum. People had to bring their own cotton wool. And I remember our neighbors, who worked at a chicken factory. When the business didn’t have cash, it gave them a salary of chickens. Numberless chickens ran all over the yard.
I left Avdiivka in 2014 when Russian-backed separatists declared east of my city to be the Donetsk People’s Republic. The roads I left on were covered with deep holes that nobody had ever come to fix, so people filled them with the remnants of the coal they burned on their stoves at home. You could lose your car in those holes. But this spring, when I left Kyiv to head back east to Avdiivka and the front line, I drove about 460 miles of perfect roads in less than 10 hours.
When I saw the hospital again, I felt as if I were in a hospital in a small Western European town. The building had been completely renovated, and lots of the equipment looked new. I got a similar feeling when I visited a local school that was damaged by shelling in 2015. It, too, had been renovated and equipped with modern computers. (Both the school and the hospital were recently almost completely destroyed by the Russians.)
Ukraine has learned how to build roads, schools and hospitals. Ukrainians have been able to travel to the European Union without a visa since 2017. And when Volodymyr Zelensky, a political outsider, was elected president in 2019 on an anticorruption, pro-European platform, the incumbent president, Petro O. Poroshenko, conceded immediately.
Of course, Ukraine still has its problems. There is still corruption. We cannot say that we are satisfied with our justice system: Our courts are not independent. Before the war, large numbers of Ukrainians left to work in Poland and other countries every year. We still have a long list of work that needs to be completed.
But these problems are not the same as those in Russia, where Mr. Putin has been more or less in charge for more than 20 years and elections are basically meaningless, where badly maintained roads crisscross the country — except when they just end — and where a person can be sentenced to prison for merely expressing an opinion, like Aleksei Gorinov, a member of a local council who was recently sentenced to seven years in prison for speaking out against the war in Ukraine.
Ten years ago Ukrainians could drink beer with Russians after the European Championship soccer matches, but we didn’t realize then that Ukraine was moving forward and Russia was moving in the opposite direction. Ukraine was trying to build a path to freedom, and Russia was building a path back to the Soviet Union with Kremlin TV and petrodollars. Eventually we grew too far apart, and something snapped.
Every day for months, I have been carrying Ukrainians who have been wounded in the fight to protect what we’ve built. Now the invaders have seen what we’ve built, too. That’s a truth that they can take home with them.
Yegor Firsov is a medic in the Ukrainian military. He was a member of the Ukrainian Parliament from 2014 to 2016.
The Times is committed to publishing a diversity of letters to the editor. We’d like to hear what you think about this or any of our articles. Here are some tips. And here’s our email: [email protected].
Follow The New York Times Opinion section on Facebook, Twitter (@NYTopinion) and Instagram.
Advertisement

source

Fallout developer announces new RPG, Wyrdsong, at Gamescom 2022 – Polygon

Filed under:
Jeff Gardiner leans on the ‘unreliable narrator’ with Wyrdsong
By now you’ve probably seen the teaser trailer for Jeff Gardiner’s next game, announced Tuesday at Gamescom, and based on his resumé you’ve might have formed some word-association impressions of what the Fallout and Elder Scrolls veteran is up to with his new studio. Terms like open-world, fantasy RPG, or lore-heavy.
Stop right there. “The lore part, that’s very interesting, because I’m actually trying to avoid lore for the most part,” said Gardiner, whose closed his 16-year tenure with Bethesda Game Studios one year ago as of Friday. “So I’m setting it somewhat historical. […] I want to explore this idea of the player being an unreliable narrator, and there’s no better setting than this, sort of semi-mythological, [Knights] Templar sort of backdrop.”
That “unreliable narrator,” familiar to those who sunk hundreds of hours into The Elder Scrolls 3: Oblivion, or Fallout 3, or any of a dozen RPGs that inspired both, is the keystone to Wyrdsong, which Gardiner is developing with Charlie Staples, a longtime friend and Obsidian Entertainment alumnus, under a new studio called Something Wicked Games. Their project has no launch window — the studio itself has only 15 employees out of the 70 that Gardiner and Staples reckon they need.
But they do have backing from NetEase, a $13.2 million seed round to be precise. And Something Wicked Games is visiting Gamescom this week for old-school, trade-convention purposes: Wave the flag, make a business deal or two, pass around some business cards and maybe get some names and phone numbers of developers who’d like to party up. That’s sort of how Gardiner himself landed at Bethesda Game Studios back in 2005, when he saw Oblivion at E3 and talked Todd Howard into a job offer.
If all of this sounds like Gardiner had a plan when he left Bethesda Softworks last August, he’d say you’re mistaken. Gardiner, who was most recently the project lead for Fallout 76, more or less left Bethesda with no options in hand; he said he needed to step away after more than a year of post-launch development during the COVID-19 pandemic. “I was lucky enough to be able to take some time off [thanks to] the Microsoft acquisition,” Gardiner said. So he didn’t really have an idea of what to do with his free time.
“COVID probably didn’t give me enough time to reflect,” Gardiner says. “I immediately started playing 100-hour RPGs; I replayed Baldur’s Gate 1, 2: Shadows of Amn, stuff like that. I realized that, while this sounds like a great idea to play 16 hours a day of RPGs, like I did when I was a kid, it’s not really that fulfilling.”
Soon enough, Gardiner was looking for work. “I did kick some tires, I pursued some executive manager roles at various studios and publishers,” he said. “I thought about consulting.” But a nostalgia for the days when he was directly involved with a game’s creative staff — that Fallout 3 to Skyrim sweet spot of 2007 to 2011 — always lingered.
The short version: Gardiner and Staples put together a pitch deck, made some phone calls, and nearly all of them were returned to set up a meeting. The two only needed the second sit-down, with NetEase, where the enormous Chinese gaming and internet conglomerate agreed to their seed-funding figure before the conversation was done. That was back in April.
“They believe in the game, they believe in the title, and they’re fully committed to supporting us,” Gardiner said. “You want to be sure you’re taking the best deal, and it was definitely NetEase. We had other offers, and we’re still going to be looking for future offers, to be honest with you, because the game is going to be a little bit bigger than the seed fund. But I think we’re well positioned for success to fund this game and get this out.”
When Gardiner describes Wyrdsong (it’s pronounced “weird-song”), it doesn’t sound like one of those auteur, statement-of-self works that a director has been busting to get out of his system for a long time — hence Gardiner’s deliberately lore-light approach. Truth be told, Gardiner’s inspiration for Wyrdsong comes from the last vacation he and his wife took before the pandemic, to Portugal. Wyrdsong is set in that country, circa 13th to 14th century, but in a preternatural world with its own laws of nature and magic.
Gardiner said after their visit to Portugal, he read Freddy Silva’s “First Templar Nation,” a 2017 thesis about the founding of the Knights Templar, a military order that has shown up in pop-culture adventures ranging from Indiana Jones and the Last Crusade to the Assassin’s Creed franchise. Silva’s book connected Gardiner’s imagination to several places he’d visited in person, and eventually his wife flat-out said he should set his next game in medieval Portugal, given how much he talked about it. Gardiner accepted the advice and — albeit after 30 revisions — created the pitch deck that would hang a shingle for Something Wicked Games.
Still, the fact Gardiner is chief executive of a new studio means he can’t completely revisit the purely creative roles he held on games more than a decade old. But he is experiencing and enjoying creative direction in a different way, through establishing a big-picture vision and then trusting colleagues to embellish the finer points and make them shine in ways he hadn’t considered.
“What I did on Skyrim, my primary focus, I was testing, tracking, following up, helping people get what they needed to be successful,” he says. “I managed people for a large part of my career, and the most important thing is just — hand it off. They have this thing called ‘Kill your darlings,’ like, just give them stuff. Let them have fun with it.
“Part of the fun of creativity is seeing where these brilliant minds can take you, not going, ‘Oh, we have to get to this place, or it’s going to be terrible,” Gardiner said. “No, it’s the opposite. They’re going to take you on this journey.”
A weekly roundup of the best things from Polygon
Please check your email to find a confirmation email, and follow the steps to confirm your humanity.
Oops. Something went wrong. Please enter a valid email and try again.
Please check your email to find a confirmation email, and follow the steps to confirm your humanity.
Oops. Something went wrong. Please enter a valid email and try again.

source

LIV Golf Bedminster tee times, TV coverage, live streams & more to watch Rounds 1-3 – Sporting News

LIV Golf is back for its third event of the season Friday, this time in Bedminster, New Jersey. After spending the end of June in Portland at Pumpkin Ridge, LIV’s field will be playing at Trump National Golf Course.
Branden Grace is the defending winner of a LIV tournament, and he’ll look to defend his title against newcomers headlined by Henrik Stenson. The teams indicate captains are starting to get comfortable with “their” guys.
It will be the first major event played at Trump’s course since the PGA Championship that was set to be played there was moved to Southern Hills in Tulsa, Oklahoma.
Here’s what to know about LIV Golf Bedminster, including start time, teams, and how to watch it live.
MORE: Why is LIV Golf playing at Donald Trump’s Bedminster course?
LIV Golf doesn’t have a full-fledged TV deal at this time and is instead hoping to build its popularity through accessibility. LIV Golf Portland, like LIV’s first tournament in London, can be watched in its entirety on YouTube, Facebook or the LIV Golf website for free around the world.
LIV Golf also prides itself on its ability to reach “over 150 territories around the world” with its free-to-air broadcast plan. Here is the complete broadcast guide by country.
LIV Golf will not be broadcast on TV in the United States. You must stream the three-day tournament to watch it, which can easily be done using the tour’s website, YouTube or Facebook pages.
Viewers in the U.K. can watch coverage on DAZN and Sky.
Once again, LIV Golf is adding to its broadcast team from event to event. David Feherty, who has been in broadcasting since 1997, will make his LIV Golf debut this weekend.
He joins Arlo White in the booth, alongside Su-Ann Heng, who provides on-course coverage. Jerry Foltz and Dom Boulet have also handled color duties alongside White.
LIV Golf will utilize a ‘shotgun start’ for its events, meaning that every player in the field will tee off at the same time. They will all begin play on different holes and complete the 18-hole course in order from there.
Play will begin around 1:15 p.m. local time in Bedminster and 9:15 p.m. in the UK. 
MORE: LIV Golf Bedminster field 2022
Each LIV Golf event will feature 12 teams that are drafted the week of the event by a tour-selected captain. Here is the breakdown of each captain and team for LIV Golf Bedminster.
Despite Grace winning the individual portion of the tournament, Aces took the team prize home thanks to dominant performances from Dustin Johnson and Patrick Reed. The team is back intact heading into New Jersey.
The LIV Golf Bedminster purse will be the customary $25 million, which is larger than any event purse on the PGA Tour. The winner receives $4 million, while the winning team gets a $750,000 apiece bonus in addition to their payout.
LIV Golf Bedminster will be played at the Old Course at Trump National Golf Course.
The controversial location will be hosting the breakout league just over a month after the PGA Championship balked at being held there and instead moved to Oklahoma.

source

Multi-temperature experiments to ease analysis of heterogeneous binder solutions by surface plasmon resonance biosensing | Scientific Reports – Nature.com

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Advertisement
Scientific Reports volume 12, Article number: 14401 (2022)
Metrics details
Surface Plasmon Resonance (SPR) biosensing is a well-established tool for the investigation of binding kinetics between a soluble species and an immobilized (bio)molecule. While robust and accurate data analysis techniques are readily available for single species, methods to exploit data collected with a solution containing multiple interactants are scarce. In a previous study, our group proposed two data analysis algorithms for (1) the precise and reliable identification of the kinetic parameters of N interactants present at different ratios in N mixtures and (2) the estimation of the composition of a given mixture, assuming that the kinetic parameters and the total concentration of all interactants are known. Here, we extend the first algorithm by reducing the number of necessary mixtures. This is achieved by conducting experiments at different temperatures. Through the Van’t Hoff and Eyring equations, identifying the kinetic and thermodynamic parameters of N binders becomes possible with M mixtures with M comprised between 2 and N and at least N/M temperatures. The second algorithm is improved by adding the total analyte concentration as a supplementary variable to be identified in an optimization routine. We validated our analysis framework experimentally with a system consisting of mixtures of low molecular weight drugs, each competing to bind to an immobilized protein. We believe that the analysis of mixtures and composition estimation could pave the way for SPR biosensing to become a bioprocess monitoring tool, on top of expanding its already substantial role in drug discovery and development.
Surface Plasmon Resonance (SPR)-based biosensors were first commercialized in the 1990s by Pharmacia. Since then, their ability to measure binding between an immobilized species and a solution species in real time without the need of any label has generated great interest in the biopharmaceutical field. Through repeated advances in the liquid handling and control systems1,2,3,4,5, the development of robust protocols6 and reproducible assays, as well as elaborate data analysis methods7,8,9,10,11,12,13,14,15, SPR biosensors have become a prominent tool for the study of biomolecular interactions and drug development.
With the recent implementation of the quality by design (QbD) and process analytical technology (PAT) frameworks16,17,18, quality monitoring tools have never been more needed. In this context, it appears appropriate to explore the potential of SPR biosensing for monitoring. Some SPR-based monitoring applications have already been reported for a variety of biotechnological applications (biotherapeutics, vaccines, bacterial detection)19. Still, to further establish SPR as a monitoring tool rather than only an investigation tool, new assays and data analysis algorithms will prove necessary.
In that endeavor, our group is interested in extending the use of SPR biosensing to the determination of the composition of biomolecule mixtures. For this purpose, a ligand that is common to all biomolecules in a mixture (the analytes) must be immobilized on the sensor surface. Then, reconstructing the signal recorded when injecting a mixture of analytes on the sensor surface, as a sum of the contribution of each analyte, should allow composition estimation. However, this requires prior knowledge of the analyte-specific kinetics for each analyte-ligand interaction. In cases where the analytes are easily purified, classical single-analyte experiments may be performed to uncover the kinetic behavior of each interaction. However, in cases where the biomolecules of interest are difficult to separate from one another (i.e., when they are only available as part of a mixture), advanced parameter identification algorithms must be employed to extract analyte-specific kinetics from multi-analyte experiments.
An example of such systems of biomolecules that form complex mixtures that are difficult to purify is the various glycoforms of therapeutic monoclonal antibodies that are produced via mammalian cell culture. The glycosylation state of these antibodies affects binding to their receptors20,21,22,23,24,25,26, as can be assessed by SPR27. Hence, advanced modeling techniques are required to analyze antibody sensorgrams and elucidate the kinetic behavior of the individual glycoforms. An SPR assay capable of monitoring the glycosylation profile of proteins would be of great interest for quality control purposes.
Apart from differences in glycosylation state, other potential applications of the multi-analyte framework include cases where molecules of different sizes or different bioactivities (due to cleavage or partial denaturation, for example) are present in the same solution and interact with the same ligand.
In a previous study15, we demonstrated that it is possible to identify the kinetic parameters of (N) analytes with the same number of independent mixtures ((M)) of the analytes (i.e. (M=N)). Furthermore, we showed that the composition of an unknown mixture could be estimated using these identified kinetic parameters. These methods were validated using experimental data obtained by immobilizing the enzyme carbonic anhydrase II (CAII) and injecting mixtures of four CAII binders. Two main limitations remained in this framework. First, the number of available mixtures might be limiting when (N) is large. Second, the composition estimation algorithm required prior knowledge of the total concentration of all analytes combined, which is not always readily available.
To address these limitations, the present study suggests two new algorithms that will be applied on the same system as in15. The first algorithm aims to reduce the number of necessary mixtures by injecting the available mixtures at different temperatures. A priori, kinetic rates and affinities of different analytes will exhibit different temperature dependencies. In brief, the information that is lacking because (M<N) is replaced by surveying at different temperatures. As the kinetics and affinities are evaluated at different temperatures, thermodynamic analyses via the Eyring and Van’t Hoff equations can be performed. These equations have previously been used to evaluate enthalpic and entropic changes via single-analyte SPR experiments with CAII binders28. The second algorithm presented here aims to estimate both the composition and the concentration of an unknown mixture of analytes, assuming prior knowledge of their kinetics. Essentially, the concentration is added as a supplementary parameter to be identified along with the analyte fractions. Only sensorgrams recorded at one temperature are necessary for our second algorithm. Therefore, we also included a guideline on how to choose the optimal injection temperature for composition estimation. These two algorithms were validated with the previously mentioned CAII system.
Biacore T100 biosensor, research-grade CM5 sensor chips, HBS-EP buffer (HEPES Buffered Saline with 30 mM EDTA [ethylenediaminetetraacetic acid] and 0.5% (v/v) surfactant P20), 70% v/v glycerol in water and ethanolamine were purchased from Cytiva (formerly GE Healthcare, Marlborough, MA). N-ethyl-N’-(3-dimethylaminopropyl) carbodiimide (EDC), N-hydroxysuccinimide (NHS), carbonic anhydrase isozyme II (CAII) purified from bovine erythrocytes, glacial acetic acid, sodium acetate, 4-carboxybenzenesulfonamide (CBS), 1,3-benzene-disulfonamide (BDS), sulfanilamide, furosemide and dimethyl sulfoxide (DMSO) were purchased from Sigma-Aldrich Canada Ltd (Oakville, ON).
The CM5 sensor chips used in this study are composed of gold on which carboxymethylated dextran has been grafted by the chip supplier. The biosensor surface preparation was performed according to a previously published protocol29. In short, two sensor chip surfaces were activated at 25 °C with an injection of 1:1 (v/v) 0.4 M EDC and 0.1 M NHS during 7 min at 20 μL/min. A solution of 0.1 g/L of CAII was prepared in a 10 mM acetate buffer at pH 5.0. This solution was injected on only one of the surfaces (experimental surface) through 30 s pulses at 20 μL/min until a density of immobilized CAII of approximately 5000 Resonance Units (RU) was reached. Next, the remaining active sites were blocked by injecting ethanolamine (1 M at pH 8.5) at 20 μL/min during 4 min. This procedure led to an immobilized CAII density of approximately 4500 RU. For each CAII surface, a mock surface (for referencing purposes) was generated by activating/deactivating the surface with the same solutions (without any CAII injections). Before and after the preparation of the surfaces, the system was extensively primed with running buffer (HBS-EP containing 3% of DMSO) and a normalization procedure with a solution of 70% (v/v) glycerol in water was performed after ligand immobilization. For more information on SPR biosensors inner-workings, we refer the reader to relevant reviews1,2,3,4,5.
HBS-EP containing 3% of DMSO (HBS-EP + 3%DMSO) was used as the running buffer. All the analytes were dissolved in DMSO and aliquoted at the following concentrations: [CBS] = 596.42 mM, [BDS] = 118.51 mM, [sulfanilamide] = 563.30 mM, [furosemide] = 3075.30 mM. For each experiment, one aliquot of each analyte was dissolved in HBS-EP + 3%DMSO to reach the following concentrations: [CBS] = 5.96 mM, [BDS] = 1.19 mM, [sulfanilamide] = 5.63 mM, [furosemide] = 30.75 mM. These stock solutions were then used in the preparation of the different dilutions and mixtures analyzed in this study.
For classical single-analyte kinetic experiments, CBS, BDS, sulfanilamide and furosemide were injected alone at 7 different concentrations (two-fold dilutions) ranging from 350 nM to 22 400 nM for CBS and furosemide, 175 nM to 11 200 nM for BDS and 900 nM to 57 600 nM for sulfanilamide. All injections were performed in duplicates and both repetitions were used separately in the analysis. The flow rate was 85 μL/min and the data collection rate was 10 Hz. Analytes were injected at four different temperatures: 12, 16, 20 and 24 °C. For double-referencing purposes, 4 buffer injections per analyte per temperature were also performed. The injection time was set to 240 s for all analytes and the dissociation time of the resulting complexes ranged from 270 to 570 s and was adjusted for each analyte and temperature such that the SPR signal reaches zero towards the end of the dissociation phase. Hence, no regeneration step was needed in between analyte injections as the dissociation was complete in every case. This agrees with various previous studies in which this system was analyzed via SPR10,11,12,15,29,30. The SPR biosensor used in this study (Biacore T100) uses the geometry proposed by Kretschmann and Raether1,3. The detection is performed by varying the incident angle while keeping the wavelength constant.
The four analytes were combined to create four random mixtures A to D, as described in Table 1. The mixture compositions were chosen randomly while ensuring that all mixtures contain all the analytes, and that no analyte has a dominating presence in any of the mixtures. All mixtures were injected at four different temperatures: 12, 16, 20 and 24 °C. In every case, the association phase duration was set to 240 s and the dissociation time was adjusted such that null responses were obtained at the end of the sensorgrams (ranging from 270 to 570 s), circumventing the need for a regeneration step. The injection flow rate was set to 85 μL/min and the data collection rate was 10 Hz. The mixtures were injected in two-fold dilutions ranging from 350 nM to 22 400 nM.
The CAII surface may wear off in time as more and more experiments are performed. This can cause a loss of activity of the immobilized ligand molecules, leading to responses with slightly lower amplitudes. To account for this loss, the sensorgrams were standardized. Before the dilution series of a given mixture was injected at a given temperature, a standard furosemide injection was performed. This was repeated for each mixture and each injection temperature. This allowed to monitor the condition of the CAII surface. A saturating concentration of furosemide was used for the standard (51 480 nM). Injections lasted 60 s, which was long enough to reach equilibrium at every temperature. The value of the plateau (standard value) was extracted. The standardization process was performed for a given sensorgram by comparing the standard value of the previous standard injection with that of the very first standard injection:
Standard injections were also double-referenced according to the procedure described in6.
Data analysis algorithms for single- and multi-analyte experiments assumed that each analyte bound to CAII according to a 1:1 Langmuir model with a constant theoretical maximal response and temperature dependent kinetic rate constants. In the case of classical single-analyte experiments, identification of the interaction kinetics was performed simultaneously for all temperatures. Thus, the maximal response of a given analyte was assumed to be constant for all temperatures. Adding this constraint did not significantly alter the identified parameters compared to fitting the model independently for each temperature: deviations of the order of 5% were observed compared to independent fits. This procedure was performed with a custom-designed script in the Matlab R2018b software platform (The Mathworks, Natick, USA). Parameters obtained with our procedure also matched those obtained with the Biacore T100 Evaluation Software31.
We developed two algorithms. The first one aims to extract the temperature dependent kinetic parameters from a set of sensorgrams corresponding to various analyte mixtures (Table 1) injected at different temperatures. This also allows the identification of thermodynamic parameters. The second algorithm estimates the composition of a given mixture and its concentration based on the prior knowledge of the kinetic (and thermodynamic) constants of each analyte. Both algorithms are presented in the next section. Fitting the multi-analyte data and estimating mixture composition was also performed using in-house Matlab scripts which are available upon request. Once the recorded sensorgrams have been retrieved from the Biacore Evaluation Software and inputted in the Matlab environment, the rest of the method is automatic, be it for parameter identification or unknown mixture analysis. As sensorgrams corresponding to the different dilutions of the different mixtures injected at different temperatures needed to be analyzed simultaneously within the fitting procedure, this type of analysis will be further described in this research as the ‘global-global-global’ approach.
Our first aim is to extract kinetic parameters ({k}_{a,i} left[=right] {s}^{-1}.{M}^{-1}) and ({k}_{d,i} left[=right] {s}^{-1}) and maximum response amplitude ({R}_{max,i}= left[=right] RU) ((i=mathrm{1,2},dots N)) of (N) analytes from (M<N) mixtures with known proportions of each analyte ({F}_{i}). To achieve this, the mixtures will be injected on a SPR surface harboring a common ligand that binds all the analytes. Injections will be performed at different temperatures. Hence, association and dissociation rates (({k}_{a,i}) and ({k}_{d,i})) will be obtained for multiple temperatures. This will allow the identification of thermodynamic parameters that describe the temperature dependence of ({k}_{a,i}) and ({k}_{d,i}) which follows the Eyring equation. Knowing ({k}_{a,i}) and ({k}_{d,i}), the affinity of the analyte-ligand interaction (({K}_{A,i})) is calculated: ({K}_{A,i}=frac{{k}_{a,i}}{{k}_{d,i}} left[=right] {M}^{-1}). Temperature dependence of ({K}_{A,i}) is described with the Van’t Hoff equation. Both the Eyring and Van’t Hoff equations were previously used with the present analyte-ligand system in a single-analyte experiment context28.
Our second aim is to use the identified parameters in order to estimate the composition of an unknown mixture of the (N) analytes, that is the proportion of each analyte in the mixture. A procedure has already been proposed by our group for the case where the total concentration of all analytes combined is known15. Here, we suggest an extension of this method for the case where the total concentration is also unknown. Ultimately, both the composition and the concentration will be determined.
We assume a Langmuir 1:1 interaction between each analyte (({A}_{i})) and the ligand ((L)):
With ({k}_{a,i}) and ({k}_{d,i}) corresponding to the association and dissociation rate constants of the ({i}^{th}) analyte, respectively. For (N) analytes, the system is described by the following ordinary differential equation (ODE) system:
({R}_{max,i}) is the theoretical SPR response that would be obtained if an infinite concentration of analyte (i) were injected. ({F}_{i}) is the fraction of analyte (i) in the injected mixture and ({C}_{TOT}) is the total concentration of all analytes. We can then compute the predicted SPR response (({R}_{TOT})) and the response caused by each analyte (({R}_{i})) by solving the system of ODE in (3).
The Eyring equation describes the evolution of a kinetic parameter ((k) standing for either ({k}_{a}) or ({k}_{d})) with respect to temperature ((T)):
(Delta {G}^{*}=Delta {H}^{*}-TDelta {S}^{*} left[=right] cal.mo{l}^{-1}) is the Gibbs energy of activation, and (Delta {H}^{*} left[=right] cal.mo{l}^{-1}) and (Delta {S}^{*} left[=right] cal.mo{l}^{-1}.{K}^{-1}) are the activation enthalpy and entropy for the pseudo-reaction describing the analyte-ligand complex formation (({k}_{a})) or dissociation (({k}_{d})). (Delta {H}^{*}) and (Delta {S}^{*}) are assumed to be temperature independent. ({R}_{IG}) is the ideal gas constant, and ({k}_{B}) and (h) are the Boltzmann and Planck constants, respectively: ({k}_{B}=1.381*{10}^{-23}J.{K}^{-1}) and (h= 6.626*{10}^{-34} J.s).
The Eyring equation can be rewritten in simpler terms for both kinetic parameters by introducing constants ({C}_{1}) through ({C}_{4}):
(c=1 {M}^{-1}) is an added factor to account for the fact that the association pseudo-reaction is not unimolecular. A priori, activation enthalpies and entropies are analyte-specific, which explains the presence of (i=mathrm{1,2},dots N) indices in (5).
The SPR response caused by analyte (i) at equilibrium, noted ({R}_{eq,i} left[=right] RU), is the amplitude of the response related to the presence of analyte (i) near the SPR surface when the response has reached a plateau during the association rate. The response of every analyte at equilibrium is given by:
The total equilibrium response of a mixture of analytes (({R}_{eq,TOT} left[=right] RU)) is analogous to the response of a single analyte with affinity ({K}_{A,obs}) and maximum response ({R}_{max,obs}):
For a given mixture and temperature, ({K}_{A,obs}) and ({R}_{max,obs}) are identified using the observed plateau values for several injections at varying concentrations. These observed parameters are then linked to the analyte-specific parameters in the following way:
We denote ({Z}_{i}) the contribution to the SPR response at equilibrium of analyte (i) (dimensionless). It represents the fraction of the total response that is caused by a given analyte:
If more than one mixture of the same analytes is available, the contribution of analyte (i) in mixture ({m}_{1}) is linked to that of the same analyte in another mixture ({m}_{2}):
If the mixtures were injected at more than one temperature, the dissociation kinetic at temperature ({T}_{1}) is linked to that of another temperature ({T}_{2}), via the Eyring Eq. (5):
For mixture (m) injected at temperature (T), if one assumes equilibrium was reached during the association phase, the SPR response during the dissociation phase (({R}_{tot}left(tright))) is mathematically described by the following sum of exponentials:
To introduce the equilibrium contributions (({Z}_{i,m})), we divide the SPR response at each time step (t) by the equilibrium plateau value ({R}_{tot}left({t}_{diss}right)) with ({t}_{diss}) indicating the beginning of the dissociation phase:
This eliminates the effect of concentration: the normalized sensorgrams (({R}_{norm}), dimensionless) corresponding to different analyte concentrations will be overlaid. A minimum of two concentrations is still necessary so that ({K}_{A,obs}) and ({R}_{max,obs}) may be identified using the equilibrium values.
The suggested analysis framework consists of a two-part algorithm that is illustrated in Fig. 1. The first part of the algorithm focuses on the dissociation phase of the recorded SPR sensorgrams. Parameters estimated in the first part will be used in the second part as a starting point for a subsequent optimization routine which takes the whole sensorgrams into account. As sensorgrams corresponding to multiple concentrations of multiple mixtures at multiple temperatures will be considered in the same fitting procedure, the fit may be called ‘global-global-global’ as opposed to ‘global’ fitting (multiple concentrations, traditional) and to ‘global-global’ fitting (multiple concentrations and mixtures15). As the multi-analyte SPR model is only locally identifiable32, providing an appropriate starting point (first part of the algorithm) is crucial to ensure convergence to a biologically meaningful solution during the second part of the algorithm, hence the pertinence of the present work.
Algorithm for the estimation of the kinetic parameters of (N) analytes at ({N}_{T}) temperatures. The first part of the algorithm surveys the dissociation phase of the sensorgrams. The results of the first part are used as a starting point for the second part which consists of a second optimization routine over the whole sensorgrams that leads to the final estimates. Estimates of ({k}_{a,i,T}), ({k}_{d,i,T}) and ({R}_{max,i}) for (i=1, dots , N) and (T=1,dots ,{N}_{T}) can be obtained from SPR sensorgrams of (M) mixtures of the (N) analytes at ({N}_{T}) temperatures ((M{N}_{T}ge N)). The fit is said ‘global-global-global’ as sensorgrams corresponding to different overall concentrations, different mixtures and different temperatures are used simultaneously in the fitting procedure. Thermodynamic parameters can be extracted from the kinetic parameters at different temperatures via the Van’t Hoff and Eyring equations.
The parameter vector (theta) contains a total of (Ncdot ({N}_{T}+2)) parameters:
The equilibrium contribution of each analyte in the first mixture (reference mixture ({m}_{1})) (({Z}_{i,{m}_{1},T})) for each temperature ((Ncdot {N}_{T}) parameters).
The dissociation rate of each analyte at the first temperature (reference temperature ({T}_{1})) (({k}_{d,i,{T}_{1}}), (N) parameters).
Temperature dependence of the dissociation rates (({C}_{4,i}), (N) parameters).
To identify the parameter vector (theta), the following conditions on the available data must be met:
Minimally (Mge 2) mixtures
(M<N), otherwise the method described in15 may be applied independently for each temperature.
The condition (frac{{F}_{i,{m}_{1}}}{{F}_{i,{m}_{2}}}ne frac{{F}_{j,{m}_{1}}}{{F}_{j,{m}_{2}}}) must be verified for each analyte couple (left(i,jright)) to validate structural identifiability i.e., the analytes fractions in different mixtures must not be perfectly correlated.
({N}_{T}) injection temperatures such that ({N}_{T}cdot Mge N).
The optimization routine aims to minimize the sum of squared residuals between the predicted normalized response and the observed normalized response. This is done for each analyte concentration, mixture and temperature, hence ‘global-global-global’ fitting.
For mixture (m) injected at temperature (T), the normalized response (independent of concentration) can be expressed in this manner (from (11), (12) and (14)):
The observed normalized response is obtained directly from the recorded sensorgrams (see (14)):
Finally, the value of the objective function (J) is given by:
As ({R}_{norm}) is not dependent on concentration, normalized responses obtained for different concentrations may be averaged to obtain ({R}_{norm}^{obs}left(tright)).
Various boundaries and constraints are added to the problem. These ensure convergence to a solution that is biologically meaningful. For example, no analyte may cause a negative response, and kinetic rates should be positive and only increase with temperature. Those are discussed in detail in Supplementary Materials, as well as a relevant starting point for the optimization routine.
After the optimization problem has been solved, we have access to (Z_{{i,m,T}}) for all analytes for each mixture and temperature. From the definition of the equilibrium contribution (10), we have:
From the definition of the observed affinity (8), we can sum on the analytes on both sides to obtain:
We obtain a (Ax=b) type linear system of equation with ({N}_{T}cdot M) equations and (N) unknowns that can be solved to obtain (frac{1}{{R}_{max,i}}) for each analyte which is inverted to obtain ({R}_{max,i}).
The resolution of this system requires to choose the number of temperatures at which the experiments are performed such that ({N}_{T}cdot Mge N). Comparing with previous work15, the information that is missing because (M<N) is compensated by surveying different temperatures.
Having access to ({Z}_{i,{m}_{1},T}) and the ({R}_{max,i}), we obtain the affinity of each analyte at each temperature from the definition of the equilibrium contributions (10):
Because the ({k}_{d,{i,T}_{1}}) and ({C}_{4,i}) are now known, the dissociation rate of each analyte can be calculated for any (T). By definition of the affinity, which was just obtained, we have:
To obtain the final estimates of the kinetic constants and response amplitudes, we feed the results of the first part of the algorithm to a second optimization routine, which takes the whole sensorgrams into account, rather than only the dissociation phase. Again, sensorgrams obtained from different mixtures injected at different concentrations and temperatures will be considered, hence we perform a ‘global-global-global’ fit. (Ncdot left(2{N}_{T}+1right)) parameters need to be identified here. Constraints may be added to ensure that kinetic rates increase with temperature. Because an appropriate starting point is provided, convergence to a biologically sensible solution should be attained. This is not necessarily the case for any starting point, as the multi-analyte SPR model is not globally structurally identifiable32.
At this stage, one can add local bulk effect contribution parameters (({R}_{I}), one per sensorgram) to account for the imperfect referencing of the buffer change effect during the association phase. Hence, the predicted sensorgram becomes:
Having access to ({k}_{a,i}), ({k}_{d,i}) and ({K}_{A,i}) at multiple temperatures, the parameters of the Eyring and Van’t Hoff equations are directly obtained via a simple linear regression. This enables the calculation of these parameters for any temperature. The Eyring equation is given in (4) and (5). The Van’t Hoff equation is given by:
where (Delta {H}^{0}) and (Delta {S}^{0}) are respectively the standard reaction enthalpy and entropy of the analyte-ligand complex formation pseudo-reaction.
Having determined the kinetic and thermodynamic parameters of the analytes, our second algorithm aims to use that information to estimate the composition of an unknown mixture of the analytes. As an extension of previous work by our group15, we will now consider the case where the concentration of the mixture is also unknown. However, we will consider that the dilution factors used to generate dilutions of a stock solution of the mixture are known.
Similarly to the parameter identification algorithm, the proposed algorithm for composition estimation also has two parts. The first one focuses on the dissociation phase of the SPR sensorgrams. The method can theoretically be used with only one sensorgram of the unknown mixture at one temperature, as long as an equilibrium plateau is reached during the association phase. Figure 2 illustrates the algorithm.
A two-part algorithm for the estimation of the composition and concentration of an unknown mixture of (N) analytes with known thermodynamic parameters. The unknown mixture is injected at temperature (T) for which corresponding kinetic parameters may be calculated using the Eyring equation. The first part of the algorithm surveys the dissociation phase of the sensorgrams to provide a first estimate of the fractions ({F}_{i}) of each analyte. The starting point for this optimization routine is set to (1/N). A first estimate of the total concentration ({C}_{TOT}) can then be found via linear regression. The second part of the algorithm considers the whole sensorgrams to provide a final estimate of ({F}_{i}) and ({C}_{TOT}). Note that, if sensorgrams of more than one total concentration are available, the corresponding dilution rates ((Dleft(sright))) must be known.
From the definition of the normalized SPR response during the dissociation phase with respect to the equilibrium contributions given in (14), we aim to minimize the following objective function via optimization:
Here, the vector of identified parameters (theta) contains the fraction of each analyte in the unknown mixture ({F}_{i}). Given a set of ({F}_{i}), the contributions ({Z}_{i}) can be calculated. Constraints should be added such that ({F}_{i}) are contained between zero and one, and that they sum to one. The starting point for each ({F}_{i}) can be set to (1/N).
Having identified the fractions, the observed affinity and maximum response can be obtained from (8) and (9). We can rearrange the expression of the equilibrium response given in (6) to obtain:
Here (Dleft(sright)) is the vector of dilution factors used to obtain the analyte concentration injected to obtain sensorgram (s). If ({C}_{TOT,REF}) is the concentration of the initial solution that was used to create a dilution series, we have ({C}_{TOT}left(sright)=Dleft(sright){{C}_{TOT}}_{REF}). At this stage, everything is known in (25) except ({C}_{TOT,REF}). It can be obtained directly if only one sensorgram is available, or via a simple regression with null intercept if multiple dilutions of the unknown mixture were injected.
The second part of the algorithm consists in using the ({F}_{i}) identified from the dissociation phase and the ({C}_{TOT,REF}) identified from the equilibrium values as the starting point of a second optimization routine that considers the entire sensorgrams (injection and dissociation phases). The same constraints are applied on the ({F}_{i}), and the objective function is:
This leads to the final estimates of ({F}_{i}) and ({C}_{TOT,REF}), from which the analyte concentration of all sensorgrams can be computed from the known (Dleft(sright)). Bulk effect contribution parameters may be added for each sensorgram as additional identified parameters in optimization problem (26).
To test our parameter identification algorithm, we immobilized CAII on SPR biosensor surfaces. It has been confirmed by SPR studies many times that CAII binds several small compounds following a simple 1:1 Langmuir scheme10,11,12,15,28,29,30. We selected four compounds amongst the known CAII binders: CBS, BDS, sulfanilamide and furosemide and we created four mixtures containing all of them (Table 1). For details on the binding mechanisms of sulfonamides to CAII, we refer to a literature review33. These binders were selected because they exhibit a wide range of kinetic and thermodynamic parameters and refractive index increments (RII). To account for the differences in RII, our model considers distinct maximum response amplitudes (({R}_{max,i})) for each analyte. The model also assumes that there is competition between the analytes (binders) and that each analyte binds CAII following a 1:1 scheme10,12,15.
To push the boundaries of pre-existing work, where four mixtures were needed to elucidate the kinetics of four analytes ((M=N)), we proposed to reduce the demand in mixtures ((M<N)) by performing experiments at different temperatures. Hence, each of the mixtures was injected at four temperatures and seven dilutions. In addition, classical experiments where the analytes were injected alone were also performed at each temperature and for each analyte. The recorded sensorgrams allowed us to identify kinetic parameters at all four temperatures, while enthalpies and entropies were derived using the Eyring and Van’t Hoff equations. All the calculated parameters are reported in Supplementary Materials Tables S1 and S2. The standard errors we computed for parameters fitted via single-analyte experiments corresponded to less than 1% of the fitted parameter in all cases and were in good agreement with those obtained with the Biacore Evaluation Software31.
We used the kinetic parameters identified with classical single-analyte experiments to evaluate the actual contribution of each analyte for each mixture at each temperature (see Fig. 3). This can be done by solving the ODEs given in (3) for each ({R}_{i}). Of interest, BDS is seen to rapidly bind to CAII, which is indicative of a high association rate. However, in every case, it is progressively replaced either by a slower binder with a higher affinity (i.e., furosemide) or by an analyte with a higher proportion in the mixture. CBS and furosemide exhibit the slowest dissociation (lowest ({k}_{d,i})) while sulfanilamide and BDS rapidly dissociate from CAII. Furosemide has the highest maximal response (({R}_{max,i})) and hence it always has a significant contribution. On the other hand, sulfanilamide has the lowest affinity for CAII amongst the studied binders (by almost one order of magnitude). This explains why its contribution is always modest to negligible, even in mixture B where it is present in a proportion of 50%. These observations will be relevant when analyzing the performance of our parameter identification algorithm.
Kinetic analysis of the contribution of CBS (red), BDS (green), sulfanilamide (blue) and furosemide (magenta) to the SPR response of each mixture (see Table 1) at each temperature. The kinetic parameters identified with classical single-analyte experiments were used to generate this figure. The summation of these signals gives the predicted signal (black). Only the responses corresponding to the maximal concentration (22 400 nM) are shown.
We first applied our parameter identification algorithm to all four combinations of three mixtures. To evaluate the loss of performance when less data is available, we repeated the procedure with all the six combinations of two mixtures. As sensorgrams corresponding to different concentrations, mixtures and temperatures are simultaneously used to fit the model, this procedure may be called ‘global-global-global’. Parameters identified with our suggested algorithm were compared to those obtained by performing classical single-analyte experiments via what we call the deviation:
where (k) denotes any fitted parameter and ({k}_{1 analyte}) is the value of the parameter as derived through a single-analyte experiment.
Kinetic parameters derived by fitting the multi-analyte model to sensorgrams of three mixtures at four temperatures are reported in Supplementary Materials Table S1. In all cases, the reference temperature ({T}_{1}) was 12 °C and the reference mixture ({m}_{1}) was the first mixture in the data set name (for part 1 of the algorithm, see Eq. (15)). 95% confidence intervals were also calculated (see Supplementary Materials). The affinity ({K}_{A}) is not directly fitted by the algorithm but is rather derived from the association and dissociation rates. Hence, its confidence interval was calculated by propagating the standard error of ({k}_{a}) and ({k}_{d}). In all cases, the amplitude of the confidence interval corresponds to less than 7% of the fitted parameter value, with association rates having the widest intervals. Association rates tend to be more impacted by experimental error, as they always multiply the analyte concentration in the model ODEs, which is dependent on the precision of the experimenter. The quality of the fits is excellent, as can be assessed by the low χ2 values corresponding to all multi-analyte fits (({chi }^{2}=sim 0.024) for all fits). The square root of χ2 corresponds to the root mean square error and should approximate the measurement noise, if the model is appropriate. Some constraints on the association rate of sulfanilamide were active in some of the fits. These constraints ensure that the kinetic rates only increase with respect to temperature. However, active constraints obscure the meaningfulness of the calculated confidence intervals of the affected parameters as well as that of the thermodynamic parameters derived from those kinetics.
We show an example of a multi-analyte fit to mixtures B-C-D in Fig. 4. A residue analysis shows no obvious trend and very small residues, hence the 1:1 competitive multi-analyte model adequately represents the data, as expected for this system. During fitting, the first and last second of the injection phase and the first second of the dissociation phase were removed to avoid artifacts hailing from buffer changes between the different phases. In 12 °C sensorgrams, a spike is always observed around 700 s. This occurs when the buffer syringe needs to be refilled due to the longer dissociation phase at this temperature.
Kinetics analysis of the injection of mixtures of four compounds at four temperatures. Black dots correspond to control-corrected and double-referenced sensorgrams. Red lines correspond to ‘global-global-global’ fits for each data set (B–C–D and A–B, see Table 1 for exact mixture compositions). The total concentration (({C}_{TOT})) of the mixtures ranged from 350 nM to 22,400 nM for all mixtures and temperatures. Here, the composition and total concentration are assumed to be known. The related residual plot is given below each sensorgram data set.
Figure 5 reports the deviation (in %) between the kinetic parameters derived from single- and multi-analyte experiments. All fits showed similarly good accuracy with deviations on the order of 10%, except for that of data set A–C–D (Fig. 5A). For this data set, estimates of the parameters of sulfanilamide had a deviation of approximately 70% on average. As observed earlier (Fig. 3), sulfanilamide had a negligible contribution to the SPR response for mixtures C and D, and a very modest one in mixture A. That is, the predicted SPR response had a weak dependence on the parameters of this analyte. Hence, the quality of the fit depended weakly on the sulfanilamide parameters, making them difficult to identify accurately using this data set. This phenomenon is avoided with all other data set thanks to the presence of mixture B, where sulfanilamide is present in a large proportion and contributed more appreciably to the SPR response. Even with the other data sets, deviations on the parameters of sulfanilamide remained higher on average (Fig. 5B). Furosemide is the analyte with the highest affinity for CAII, and thus the highest contribution to the SPR response. As such, it was also the analyte for which the kinetic parameters are the most accurately identified with less than 5% deviations on average. The association rates were the parameters with the highest average deviation (Fig. 5C). As previously mentioned, this is because they always multiply the concentration (and fraction) in the ODEs, which is dependent on the experimenter’s repeatability when creating the mixtures and performing the necessary dilutions from one experiment to another. Moreover, the association rates only directly affect the injection phase of the sensorgrams. On the contrary, the ({R}_{max,i}) were most precisely identified. As these are considered temperature invariant, they have a stronger impact on the quality of the fit than any single-temperature association or dissociation rate. Hence, the fit should be more sensitive to the ({R}_{max,i}). Dissociation rates were also accurately identified, as they affect both phases of the sensorgrams. Affinities were obtained by computing the ratio of the association and dissociation rates rather than being directly identified, hence the precision of the predicted affinities is dependent on the precision of the predicted kinetic rates. Finally, kinetic parameters corresponding to all temperatures were identified with a similar accuracy (Fig. 5D).
Average deviation between the parameters (({k}_{a}), ({k}_{d}), ({K}_{A}) and ({R}_{max})) identified from single-analyte experiments and parameters identified by our multi-analyte model with any combination of three mixtures (out of the four mixtures described in Table 1). In all cases, data generated at four temperatures were taken into account. (A) Average deviation computed over all parameters and temperatures for all analytes in all fits. (B) Average deviation over all parameters, fits and temperatures for each analyte. (C) Average deviation for all parameters. (D) Average deviation for all temperatures. Error bars were computed by propagating the standard error of the parameters identified from single-analyte and multi-analyte experiments.
Having derived the association and dissociation rates, and the affinity at four different temperatures, we identified the Eyring and Van’t Hoff equation parameters via linear regression. Table S2 in Supplementary Materials reports the activation and standard reaction enthalpies and entropies of the analyte-ligand complex formation pseudo-reaction. Considering the standard Gibbs energy (Delta {G}^{0}=Delta {H}^{0}-TDelta {S}^{0}), all analyte complex formations were favored by enthalpic contributions (negative enthalpy change), whereas CAII binding to BDS or furosemide was also supported by entropic contributions (positive entropy change).
Figure 6 compares the thermodynamic parameters derived from multi-analyte experiments to those derived from single-analyte experiments. The thermodynamic parameters of sulfanilamide exhibit the biggest deviations in all fits, as expected because of its low contribution to the SPR response (Fig. 6A). Data set B–C–D produced a fit with no active constraint, leading to the most precise identification of the thermodynamic parameters (35% deviation for sulfanilamide and below 25% for all other analytes). On the other hand, fit A-B-D had the largest number of active constraints, leading to the largest deviations. The constraints that are active are those that ensure that the association rates only increase with temperature. This in turn affects the computation of the corresponding enthalpic and entropic changes.
Average deviation between the thermodynamic parameters ((Delta {H}_{{k}_{a}}^{*}), (Delta {S}_{{k}_{a}}^{*}), (Delta {H}_{{k}_{d}}^{*}), (Delta {S}_{{k}_{d}}^{*}), (Delta {H}_{{K}_{A}}^{0}) and (Delta {S}_{{K}_{A}}^{0})) derived from single-analyte experiments and parameters derived by fitting the multi-analyte model at four temperatures with any combination of three mixtures (A–D). (A) Average deviation computed over all parameters for all analytes in all fits. (B) Average deviation computed over all fits for all parameters of all analytes.
Enthalpies were accurately identified (average deviation of ~ 13% between single- and multi-analyte experiments averaged over Eyring and Van’t Hoff parameters found with all data sets) whereas entropy estimates were less accurate (average deviation of ~ 59%), as is shown in Fig. 6B. The enthalpies modulate the temperature dependence of the kinetic parameters (or affinity) while the entropies represent the base level, which may be more sensitive to experimental error during sample dilution. All predicted enthalpies and entropies had the correct sign except for entropies associated with the dissociation rate ((Delta {S}_{{k}_{d}}^{*})), mainly for sulfanilamide. This is presumably because the absolute value of this entropy seems to be close to 0 for sulfanilamide. Our algorithm can provide at least coarse first estimates when analyzing a mixture of analytes that is difficult to purify, but likely cannot compete with specialized methods, like calorimetry, in accuracy when pure samples are available.
Part 1 of our algorithm favors convergence to a proper, biologically relevant solution when performing part 2. This is necessary because the multi-analyte model is only locally identifiable32. To demonstrate this point, we performed the optimization routine of part 2 of our algorithm with a suboptimal starting point for data set B–C–D. The suboptimal starting point was constructed by assigning the value of the expected order of magnitude of each parameter (104 s−1 M−1 for ({k}_{a}), 10–2 s−1 for ({k}_{d}) and 10 for ({R}_{max})). With this data set, the optimizer converged to a suboptimal solution (χ2 of 0.0301 vs 0.0247) that was further from single-analyte estimates as the one obtained with our algorithm (average deviation over all analytes, parameters and temperatures of 74% vs 10%). Moreover, convergence necessitated approximately three times as many iterations when using a suboptimal starting point. The fitted parameters obtained with a suboptimal starting point are given in Supplementary Materials Table S3.
Parameters identified by fitting the multi-analyte model at four temperatures for any combination of two mixtures (instead of three, as in the above sections) are reported in Supplementary Materials Table S4. Confidence interval lengths and χ2 values were similar to those obtained when using three mixtures. An example of fit is shown in Fig. 4 for data set A–B. Once again, no obvious trend was observed in the residues and the model adequately depicted the data.
Figure 7 shows the deviation (in %) between the kinetic parameters obtained with multi-analyte fits with two mixtures and those obtained with single-analyte experiments. A similar accuracy was obtained with all data sets except for data set C–D (Fig. 7A). With this data set, deviations on sulfanilamide parameters were over 100% on average (a more than two-fold difference from single-analyte experiments). This is mainly because sulfanilamide had a negligible contribution to the SPR response in both mixtures. Interestingly, the other three analytes were properly identified even in the C–D data set, leading to similar deviations as those obtained with three mixtures (Fig. 7B). Average deviations for sulfanilamide were larger when fitting to two mixtures compared to three (~ 50% vs ~ 30% if we exclude data set C–D). Dissociation rates and maximal responses were again the most precisely identified parameters, while association rates and affinities were more difficult to extract (Fig. 7C), as was the case for three-mixture data sets. Parameters corresponding to each temperature were identified with approximately the same precision (Fig. 7D). Overall, the parameter estimation was more precise with three-mixture data sets with an average deviation of ~ 13% than with two-mixture data sets with an average of ~ 21% (Fig. 7E).
Deviation between the parameters (({k}_{a}), ({k}_{d}), ({K}_{A}) and ({R}_{max})) identified from single-analyte experiments and parameters identified by fitting the multi-analyte model at four temperatures with any combination of two mixtures (out of the four described in Table 1). (A) Average deviation computed over all parameters and temperatures for all analytes in all fits. (B) Average deviation for all analytes over all parameters, fits and temperatures. Average deviations for sulfanilamide are also reported without considering data set (C–D) (sulfanilamide*). (C) Average deviation for all parameters. (D) Average deviation for all temperatures. (E) Average deviation for data sets with two and three mixtures. Error bars were computed by propagating the standard error of the parameters identified from single-analyte and multi-analyte experiments.
Thermodynamic parameters derived from two-mixture data sets are shown in Table S5 in the Supplementary Materials. Again, the thermodynamic parameters of sulfanilamide differed the most from those obtained with single-analyte experiments. Except for data set C–D, the correct sign was found for the enthalpies and the average deviation (over Eyring and Van’t Hoff enthalpies found with all data sets except C–D) was ~ 18%. Entropies were identified less precisely, with an average deviation (excluding C-D) of ~ 71%. Constraints on the association rates of CBS and sulfanilamide were active in the final solution of data set C–D. There was also an active constraint on the affinity of sulfanilamide to ensure that it varies monotonously with respect to temperature. This contributed to the larger deviations observed for this data set. For more details on the deviation values, we refer the reader to Supplementary Materials Figure S1.
Before injecting an unknown mixture on the biosensor surface, the injection temperature must be determined. Having access to the kinetic parameters for theoretically any temperature within the operating range of the biosensor, after applying the parameter identification algorithm, this question is nontrivial. Selecting the temperature for which the kinetic parameters of the different analytes are most different from one another would lead to bigger differences in their binding behavior, facilitating the discrimination of the respective contributions to the SPR response. As the first part of our algorithm focuses on the dissociation phase of the sensorgrams, we hypothesized that selecting the temperature which maximizes the differences in the dissociation rates should lead to better estimates. To quantify how different the dissociation rates are for a given temperature, we suggest the following temperature dependent indicator:
The indicator (Ileft(Tright)) is equal to or greater than one. A temperature with a small (I) means that at least two dissociation rates are close to each other, whereas a large value of the indicator signifies that the two closest dissociation rates are rather far from one another.
We tested our hypothesis via simulations. The four analytes were considered, with their Eyring parameters derived from the single-analyte experiments. 100 mixtures were constructed by randomly generating sets of fractions ({F}_{i}). 15 equally distanced temperatures between 12 °C and 40 °C were selected. For each temperature and mixture, sensorgrams were simulated by solving the ODEs in (3) for two-fold dilutions ranging from 400 nM to 25 600 nM. Data were generated with a frequency of 10 Hz. Gaussian noise with a null mean was added on top of the simulated sensorgrams to reproduce the typical measurement noise observed on recorded sensorgrams. The choice of Gaussian distribution for the noise was validated via a quantile–quantile plot showing the quantiles of the double-referenced baseline signal of the instrument with respect to the theoretical quantile values of a normal distribution. A standard error of 0.2 was used, which is similar to that observed for said baseline. We applied our composition estimation algorithm to each mixture and each temperature individually. We then compared the estimated composition after the first and second parts of the algorithm to the composition used to generate the data. The absolute value of the difference between the estimated and actual fractions was recorded to evaluate the quality of the estimates, i.e., the mean absolute error (MAE):
For the concentration, we recorded the ratio of the predicted value and the actual concentration: (frac{{C}_{TOT,pred}}{{C}_{TOT,real}}). Results of the simulations are reported in Fig. 8.
Application of the composition and concentration estimation algorithm to simulated mixtures of the four analytes at 15 temperatures between 12 and 40 °C. (A) Dissociation rates of each analyte with respect to temperature. (B) Performance indicator with respect to temperature. (C) Absolute values of the fraction estimation error averaged over all analytes and mixtures with respect to the indicator after part 1 and part 2 of the algorithm. To help interpretation, corresponding temperatures were added as annotations. (D) Ratio of estimated and actual concentrations with respect to the indicator after both parts of the algorithm. The black line corresponds to a perfect estimate (ratio of 1).
The dissociation rates of BDS and sulfanilamide and that of CBS and furosemide overlap at 25 °C and 27 °C, respectively (Fig. 8A), where the value of our suggested indicator reaches a minimum, as seen on Fig. 8B. If only part 1 of the composition estimation algorithm is performed (Fig. 8C), the absolute value of the difference between the estimated and actual fractions (averaged over the 100 simulated mixtures and 4 analytes) is negatively correlated to the indicator. That is, part 1 of the algorithm functions better when the dissociation rates are more different from one another, which was expected. There is however another factor in play, which is the injection temperature. Estimation errors tend to increase with temperature. This effect is even more noticeable in the results of part 2 of the algorithm (Fig. 8C). No obvious trend was observed when we replaced ({k}_{d}) in the definition of (I(T)) with any of ({k}_{a}), ({K}_{A}) or ({K}_{D}=1/{K}_{A}).
The same pattern is also obtained in Fig. 8D, where the predicted concentration is compared to the actual concentration by taking their ratio. This was averaged over the 100 mixtures. The concentrations estimated by part 2 of the algorithm were within 5% of the actual fractions.
As the temperature increases, kinetics become faster and sensorgrams exhibit less curvature (more of a ‘rectangular’ shape). These ‘rectangular’ sensorgrams do not contain a sufficient amount of information to be able to discriminate between the analytes.
In general, we would advise to select the temperature that maximizes our indicator (I(T)), while minimizing temperature, as much as possible so that equilibrium plateaus may still be reached. This will ensure part 1 of the algorithm provides an appropriate starting point for part 2, allowing faster convergence to a better estimate of the composition.
In the simulation, sulfanilamide was the analyte with the biggest estimation errors, due to its lower affinity for CAII, and hence smaller contribution to the simulated response. BDS followed, as BDS and Sulfanilamide exhibit similar dissociation rates, complicating their discrimination. Overall, the estimates obtained after part 2 were of great quality with small errors. In absolute value, a difference of 1% was observed between the estimated and actual proportions at 12 °C. At 40 °C, the difference was of the order of 10%.
Next, we used the kinetic parameters sets identified with real sensorgrams (Tables S1 and S4 in Supplementary Materials) to estimate the composition and concentration of all eight mixtures described in Table 1. This was performed independently for every parameter set and every temperature. The algorithm detailed in the “Composition estimation” section was used. An example is shown in Fig. 9A,B. The kinetic parameters identified from data set B–C–D were used here, and only sensorgrams at 12 °C were fed to the algorithm. The vertical error bars in these plots correspond to asymmetric confidence intervals found with a method described in Supplementary Materials. The horizontal error bars were found by propagating the systematic error of the pipettes (Pipetman Neo P10, P20, P200 and P1000) and the balance (Dever Instrument Company AA-160) that were used to conduct the experiments. The small vertical error bars are mainly due to the large number of data points that are considered (injections of 7 concentrations, in duplicate with 10 points per second). For the most part, points are well aligned with the central line in these plots, indicating proper estimates. The experimental data corresponding to pure sulfanilamide were the most difficult to elucidate. This is mainly because the identified kinetic parameters of sulfanilamide exhibited the largest deviations from single-analyte estimates for this data set and temperature (~ 14% for the association rate and ~ 18% for the dissociation rate). The kinetic parameters of CBS, BDS and furosemide are relatively well identified (deviations between 0 and 11%). So are maximal responses (deviations below 10% for all four analytes). It appears that imprecisions in the identified kinetics of sulfanilamide are the driving force of the estimation errors. The fraction of sulfanilamide is overestimated in mixture A and pure CBS, leading to higher concentration estimates for these mixtures to compensate for sulfanilamide’s lower affinity for CAII. The other 5 mixtures were well identified. It might not be surprising that mixtures B, C and D were correctly elucidated, as those were used to identify the kinetics in this case study. BDS and furosemide have high affinities and their parameters were precisely identified, hence pure BDS and furosemide were also precisely estimated. BDS is falsely detected in the 100% sulfanilamide mixture, possibly because sulfanilamide and BDS have similar dissociation rates.
Performance of the composition and concentration estimation algorithm. (A) Calculated fractions with respect to actual fractions of the four analytes in each of the 8 mixtures detailed in Table 1. These fractions were estimated from sensorgrams at 12 °C only with kinetic parameters identified from data set (B–C–D). (B) Calculated concentration with respect to actual concentration for each of the 8 mixtures. These were estimated along the fractions using sensorgrams at 12 °C only with kinetic parameters identified from data set (B–C–D). (C) Mean absolute error of estimated fractions for each analyte at each temperature. The composition estimation algorithm was used independently for each set of kinetic parameters corresponding to each data set (all combinations of 2 and 3 mixtures) and each temperature, and mean absolute errors were averaged across all data sets and mixtures. (D) Mean absolute error (averaged across all data sets) with respect to the performance indicator. Annotations on the graph indicate the corresponding temperature. (E) Deviation (in %) between estimated concentrations and actual concentrations (averaged across all data sets) with respect to temperature. (F) Deviation of the concentration and mean absolute error of the fractions with respect to the data set that was used to identify the kinetics. Those were taken at 12 °C only. In this figure, ‘singles’ refers to single-analyte fits.
This corroborates Fig. 9C where sulfanilamide and BDS are shown to be the two analytes with the largest MAEs (averaged over all sets of kinetic parameters), no matter which injection temperature is considered. CBS and furosemide have much lower MAEs, with a weaker temperature dependence. As predicted by the simulations, we observed that performing the experiments at a lower temperature leads to more precise composition estimations. This cannot be explained by better kinetic parameters estimates at lower temperatures, because Figs. 5D and 7D showed no clear trend in parameter deviation with respect to temperature. Figure 9D shows that the MAE was also correlated to our suggested indicator. The same trend is observed for concentration estimates (Fig. 9E). This is easy to understand, as large errors in the fraction estimates can be somewhat compensated by the concentration estimation, and vice versa, as both the fractions and concentration are estimated concurrently.
Interestingly, the kinetic parameter sets identified from data sets A–C–D and C–D did not lead to larger estimation error. These data sets were highlighted during the parameter identification process, as they led to noticeably bigger deviations for the kinetic parameters of sulfanilamide. The other three analytes were well identified, however, and overall, these parameter sets can be used to estimate the composition and concentration with similar accuracy to other parameter sets. This is the case at 12 °C, as shown in Fig. 9F, and for other temperatures. The quality of the kinetic parameter estimates still plays an important part, as using the parameters identified from single-analyte experiments to estimate the composition and concentration led to the smallest errors (‘singles’ data set in Fig. 9F). The maximal response deviations from data set A–B were the largest. This led to large concentration estimation errors for mixtures aside from A and B. This data set led to the biggest deviations on the kinetics of furosemide, which was properly identified for all other data sets.
100% BDS and 100% sulfanilamide were the mixtures for which the estimation errors were the biggest on average. As previously mentioned, this is probably due to their similar dissociation rates, and sulfanilamide’s lower affinity for CAII.
Adding the concentration to the list of parameters to be identified adds a degree of freedom which can confound the estimation of the fractions. To evaluate this effect, we repeated the whole procedure, this time by constraining the value of the concentration to the correct experimental value. This ends up being practically the same as the algorithm proposed in15 for the case where concentration is known. The results are presented in detail in Supplementary Materials Figure S2. Fraction MAEs were approximately two-fold lower when the concentration is known (~ 4% vs ~ 8%). Removing a parameter to be fitted increases the precision of the optimization algorithm, resulting in smaller confidence intervals. Providing information to the algorithm facilitates convergence to the correct fractions as long as the set of kinetic parameters used is adequate. Larger MAEs were obtained with data set C–D when the concentrations are known, mainly driven by sulfanilamide whose parameters presented high deviations in this fit, which could not be compensated for by the concentration estimate.
Measuring the total concentration of all analytes prior to composition estimation (through an SPR experiment or otherwise), if possible, leads to more precise composition estimates. To do so via SPR, it is necessary that the analytes have the same refractive index increment. For proteins, this would mean similar molecular weights (for example different glycosylation or other post-translation-modifications of a same protein). One would first produce a calibration curve of the initial slope of a mass transport limited sensorgram with respect to the concentration of analyte using an analyte mixture of known concentration. The concentration of an unknown mixture could then be deduced from a mass transport limited sensorgram of the unknown mixture19,34,35,36.
The previous sections showed promising results for the identification of kinetic parameters of (N) analytes with (M<N) mixtures and for the estimation of the composition and the concentration of an unknown mixture of these analytes. However, in both cases, sulfanilamide is highlighted as problematic due to its lower affinity to CAII (by approximately an order of magnitude compared to the other three analytes) that makes its contribution to the SPR response marginal unless it occupies a dominating proportion of the mixture. It also exhibits a similar dissociation rate as that of BDS, leading to confounding these two analytes when estimating mixture compositions.
When one or more analytes have weak affinities and/or similar kinetics leading to poorer fraction estimates, one solution is to construct analyte pools by regrouping two or more analytes into one. This can potentially lead to more relevant estimates for the pooled analyte, as the pool might have a more appreciable contribution than each of the individual pooled analytes. This might not affect the identification of other non-pooled analytes which already have high affinities and contributions to the SPR response.
Applying this concept in our case study, we decided to pool sulfanilamide with BDS. This reduces the number of analytes to three. To remain within the scope of the parameter identification algorithm proposed in this study ((1<M<N)), we should have analyzed data sets containing two mixtures. We chose data set C–D, as both of these mixtures contained a small fraction of sulfanilamide, leading to big deviations in the identified kinetic parameters for this analyte. We used our parameter identification algorithm while considering ({F}_{pool}={F}_{BDS}+{F}_{sulfanilamide}). Results are reported in Fig. 10A and in Supplementary Materials Table S6. The kinetics of CBS and furosemide were properly identified, and the kinetics of the pooled analyte closely matched those of BDS. This is not surprising, since the contribution of the pool to the SPR response is almost exclusively due to BDS. Interestingly, the deviations on the kinetics and affinity of CBS were lower with the pooled analyte than without it and they were similar for furosemide. Maximal responses exhibited bigger deviations when pooling, but they remained well below 10%. This suggests that pooling some analytes does not prevent proper identification of the non-pooled analytes. The χ2 with pooling was slightly higher than without (0.0249 vs 0.0241) as removing parameters necessarily leads to bigger residues. However, they exhibited no obvious trend and the model still properly depicted the data.
Performance of the kinetics identification and composition and concentration estimation algorithms when pooling BDS and sulfanilamide. (A) Deviation between the parameters (({k}_{a}), ({k}_{d}), ({K}_{A}) and ({R}_{max})) identified from single-analyte experiments and parameters identified by fitting the multi-analyte model at four temperatures with data set (C–D) for CBS and furosemide. For these fits, the total concentration and fractions of all analytes are assumed to be known (with ({F}_{pool}={F}_{BDS}+{F}_{sulfanilamide})). Deviations were averaged over all temperatures. Error bars were computed by propagating the standard error of the parameters identified from single-analyte and multi-analyte experiments. (B) Calculated fractions with respect to actual fractions of the three analytes in each of the 8 mixtures. These fractions were estimated from sensorgrams at 12 °C only. (C) Calculated concentration with respect to actual concentration for each of the 8 mixtures. These were estimated along the fractions using sensorgrams at 12 °C only. (D) Mean absolute error of estimated fractions for each analyte at each temperature. The composition estimation algorithm was used independently with pooling of BDS and sulfanilamide and without pooling. Averages over all 8 mixtures are reported.
We then used this new set of identified kinetic parameters to estimate the composition of the 8 mixtures independently for each temperature. Example results are shown in Fig. 10B,C for an injection temperature of 12 °C.
The MAEs on fractions were lower when pooling. Of interest, the MAEs relative to CBS were slightly lower with pooling while those of furosemide were similar (see Fig. 10D). Deviations in the estimated concentrations were almost identical with and without pooling. The mixture with the highest concentration deviation was 100% sulfanilamide. This mixture is appropriately estimated to be composed of the pooled analyte almost exclusively (95% pool + 5% furosemide). However, since the pooled analyte’s affinity resembles that of BDS and not that of sulfanilamide, the estimated concentration is found to be almost five times lower than the real concentration. Without pooling, the mixture is estimated to be composed of (75% BDS and 25% sulfanilamide), also leading to a fivefold lower concentration estimate.
In summary, pooling a low affinity analyte and a strong affinity analyte with a similar dissociation rate made the identification of the other analytes equally or more precise while enabling better composition estimates. The kinetic parameter identification may not work as well if the data set used to fit the multi-analyte model is composed of mixtures in which the low affinity analyte has a dominating proportion. For example, repeating the procedure with data set A–B did not lead to appreciable gains in parameter or composition estimation performance. In these mixtures, sulfanilamide is present in high proportions (respectively 30% and 50%) and has an appreciable contribution to the SPR response.
This study pertains to the analysis of analyte mixtures by SPR biosensing. Those are ill-suited to traditional single-analyte studies, as analyte heterogeneity complexifies the recorded SPR signals and renders the use of a single Langmuir binding isotherm inadequate for data analysis. Using a kinetic model that properly describes the actual system is essential if one aims to uncover meaningful information on the interaction behavior. Moreover, the ability to analyze multi-analyte samples enables mixture composition estimation, which could prove a powerful tool to streamline bioprocess monitoring and development.
Carbonic anhydrase II (CAII) was used as the immobilized ligand in our experiments because of its well-known ability to form a simple 1:1 interaction with several small compounds with varying kinetic rates and affinities. Amongst CAII binders, we selected (N=4) analytes: CBS, BDS, sulfanilamide and furosemide. To simulate a realistic situation, we constructed four mixtures of these analytes spanning a wide range of compositions (Table 1). The SPR response recorded when injecting each mixture on a CAII-immobilized surface should follow a multi-analyte 1:1 binding model, where every analyte competes to bind CAII. In this paper, we suggest two algorithms. The first one aims to identify the kinetic parameters of (N) analytes with (Mle N) mixtures of the analytes. To accomplish this, the mixtures must be injected at a minimum of (N/M) temperatures, and the composition and concentration of the mixtures must be known (Fig. 1). The second algorithm seeks to estimate the composition and the concentration of a mixture, given that the kinetics have been identified previously (Fig. 2). If more than one concentration is injected, the dilution rate used to create the dilution series must be known.
We chose four injection temperatures: 12, 16, 20 and 24 °C as they allow full dissociation within a reasonable time while still exhibiting sufficient curvature during the association phase. We first injected pure samples of the four analytes at each temperature to obtain standard values of their actual kinetics. This allowed us to predict the contribution of each analyte in the four mixtures by solving the multi-analyte model ODEs (Fig. 3). We then injected the mixtures and utilized our first algorithm to identify the kinetics of the four analytes with three mixtures (Table S1). Fits to real sensorgram data were shown to be excellent (Fig. 4), and the identified kinetic parameters closely matched those obtained via single-analyte experiments, especially when sulfanilamide has an appreciable contribution to the SPR response (Fig. 5). Having identified the kinetics and affinities at each temperature, thermodynamic modeling became possible via the Eyring and Van’t Hoff equations (Fig. 6 and Table S2). The procedure was then repeated with combinations of only two mixtures (Fig. 7 and Table S4). The identified parameters still closely matched those from single-analyte experiments, but with slightly larger deviations than when using three mixtures to fit the model.
We then focused on our second algorithm. We showed, via simulations, that the algorithm performed better when the unknown mixture is injected at lower temperatures, or at a temperature for which the dissociation rates of the analytes are more disparate (Fig. 8). We then applied our second algorithm on real sensorgram data and obtained relatively precise estimates using 12 °C sensorgrams (Fig. 9). Throughout our study, sulfanilamide was highlighted as problematic because of its low affinity to CAII, leading to small contributions to the SPR response and a greater difficulty identifying its kinetics or estimating its fraction in an unknown mixture. It also has a very similar dissociation rate to that of BDS, potentially leading to confounding both analytes. We showed that considering them both as only one analyte (i.e., pooling them) could lead to a better performance of both of our algorithms, as long as the mixtures used to identify the kinetics contain little sulfanilamide (Fig. 10).
The merit of this current study compared to previous work on mixture analysis via SPR15 lies in the ability to reduce the number of mixtures required to extract the kinetics ((Mle N) instead of (M=N)). When working with mixtures of compounds that are difficult to separate, it might be difficult to obtain multiple, different and linearly independent mixtures. On the other hand, injecting the available mixtures at several temperatures is a standard feature on most SPR biosensors. The only downside being longer SPR experiments due to stabilization periods required between temperature changes. Moreover, we showed that it is possible to estimate the concentration along the mixture composition, which was never reported before.
The multi-analyte kinetic model is structurally identifiable only locally rather than globally32. Briefly, this means that several (but a finite number of) sets of kinetic parameters may lead to the same predicted SPR response. To select a biologically meaningful solution, we proposed a preliminary subroutine (part 1 of our parameter identification algorithm) to provide adequate starting points for the optimization over the whole sensorgrams (part 2 of the algorithm). This is significant, since we have shown that a poor starting point may lead to convergence towards a solution that is further away from kinetic parameters derived from single-analyte experiments in the “Importance of the Initial Estimates for Parameter Identification” section and Supplementary Materials Table S3.
The structural identifiability of part 1 of the parameter identification algorithm was briefly discussed in “Required data” section. Because summation (here of exponential decays) is commutative, at least two mixtures are necessary so that the (M) by (N) matrix of ({Z}_{i,m}) and (N) by (1) vector of ({k}_{d,i}) describing the dissociation phase of all mixtures are ordered the same way (in terms of the analytes) as the (M) by (N) matrix of analyte fractions ({F}_{i,m}). The latter is assumed to be known and is fed to the algorithm. Additionally, no analyte couple should have perfectly correlated fractions in the available mixtures. If this condition is not met, the product of ({K}_{A,i}) and ({R}_{max,i}) could be adjusted so that different orders could lead to the same set of ({Z}_{i}), obfuscating the identification of ({K}_{A,i}) and ({R}_{max,i}).
Another important concept to consider is practical identifiability. A non-identifiable model that can be made identifiable by adding data or reducing measurement noise is deemed practically non-identifiable. While there is no universal method to test practical identifiability, an elegant one consists in confirming that no fitted parameter has infinite confidence intervals. This can be done, for example, by disturbing the optimal solution one parameter at a time and re-optimizing, which leads to a profile likelihood37,38. We used the same method we used previously to derive confidence intervals on the parameters derived from part 1 of the algorithm (as detailed in Supplementary Materials), and found finite confidence intervals for all parameters, be it for data sets consisting of three or two mixtures (Fig. S3 in Supplementary Materials). Hence, the model of part 1 of the algorithm is deemed practically identifiable with the available amount of data and the noise level of our SPR instrument.
The goal of this study was to establish a proof of concept for both algorithms, not to optimize their use. Indeed, the parameter identification algorithm could theoretically work with only two injected concentrations (just enough to identify the observed affinity and maximal response) while seven were injected. The duration of the association and dissociation phases could also be shortened, as long as an equilibrium plateau is still reached and the response goes back to zero at the end of the sensorgram, so as to avoid the need for a regeneration step (in our experimental system). Some previous studies have aimed to optimize these parameters online for single-analyte experiments9,11. Furthermore, in our case study where (N=4) and (M=3) (or (M=2)), only two temperatures were theoretically necessary (({N}_{T}ge N/M)). This is dictated by the system of linear equations in (19), which needs to be solved to obtain a first estimate of the maximal responses. We recommend using more temperatures than the minimum required to help counter the effects of experimental error and measurement noise, and thus obtain an accurate solution of Eq. (19). Moreover, the selected injection temperatures should be spaced enough so that the temperature-driven differences in binding behavior are more pronounced than the experimental error and noise, again to facilitate solving Eq. (19). The same reasoning can be applied to the number of mixtures, with more mixtures leading to more accurate parameter estimates, as we observed in this study.
Low temperatures slow the kinetics, which may prevent reaching an equilibrium state in a reasonable injection time, whereas high temperatures may result in ‘rectangular’ sensorgrams which contain little information in their curvature. Hence, extreme temperatures should be avoided. The selection of temperatures to perform parameter identification will be dependent on the studied binders but beginning towards the middle of the temperature range allowable by the SPR biosensor and making positive and negative steps of 4 or 5 degrees may be a good general starting point.
The experimental error was estimated by performing the same dilution series of the same analyte three times, injecting them on the SPR surface, and identifying the resulting kinetic parameters. Deviations of up to 10% were obtained between such triplicates. Our parameter identification algorithm led to deviations of this scale when using three mixtures, whereas only using two mixtures led to deviations that are twice as high. Such seems to be the cost of reducing the amount of available data.
Our model considers a specific maximal response (({R}_{max,i})) for all analytes. This was necessary for our system, as small molecular weight compounds, such as those used as analytes in this study, exhibit analyte-specific refractive index increments12. For high molecular weight compounds, such as proteins, ({R}_{max,i}) should be linearly proportional to the molecular weight39,40,41. This would simplify the parameter identification algorithm, as only one ({R}_{max}) would need to be included, provided that molecular weights are known. For proteins, a very small temperature dependence of the refractive index increment has been reported42,43,44. However, it has been modelled using the same temperature-dependent function for multiple proteins between 10 and 25 °C42,43. Hence, the standardization procedure detailed in section “Furosemide standard injections” should allow the temperature invariant ({R}_{max,i}) assumption.
Using our algorithm could save experimental time when performing thermodynamic analyses. Indeed, rather than injecting all (N) analytes at multiple temperatures, one could inject (M<N) mixtures instead. In both cases, kinetic parameters and affinities could be identified at all temperatures, enabling the construction of Eyring and Van’t Hoff plots from which entropic and enthalpic changes could be extracted. Incidentally, the Eyring and Van’t Hoff plots obtained with single-analyte experiments showed high ({R}^{2}) ((>99%)), indicating that the hypothesis of temperature independent entropic and enthalpic changes is appropriate for our system.
For the unknown mixture composition and concentration determination, sensorgrams obtained at only one temperature were used. Using sensorgrams from more than one temperature might lead to more accurate estimates, but it seems ill-fitting to the goal of this algorithm, which is to rapidly determine the composition. Indeed, changing the injection temperature requires the biosensor to restabilize, which takes at least one hour for a precise and stable temperature. It seems more realistic to envision the case where a user would inject multiple unknown samples subsequently at the same temperature. For the parameter identification, one must perform experiments at multiple temperatures, meaning restabilizing the instrument multiple times. This process could prove lengthy, but once the parameters have been identified, they can be used to analyze a large number of unknown mixtures more promptly afterwards.
Identifying specific binding kinetics may be daunting when characterizing macromolecules that are difficult to purify. For instance, it is well documented that immunoglobulin G (IgG) production in bioreactor leads to a heterogeneous distribution of IgG glycoforms25. The identity of the glycan groups added to the IgG Fc region greatly influences therapeutic efficacy, mainly by affecting binding to Fcγ receptors present on effector immune cells20,21,22,23,24,25,26. The IgG-FcγR interaction can be measured by SPR biosensing27. The abundance of each glycoform can be deduced from mass spectrometry analysis, for example20,21,22,23,24,25,26. Hence, an algorithm similar to that proposed in15 could be used to uncover glycoform-specific kinetics. However, the number of different glycoforms is usually quite large, and while different mixtures of glycoforms can be created via bioengineering methods25, by varying cell culture conditions45 and by using different cell lines46, reducing the number of mixtures needed remains necessary. Using our approach, this could be achieved by injecting the available mixtures at different temperatures, or by pooling glycoforms that have a lower affinity for the FcγR. Having uncovered the kinetics, our second algorithm could be used to estimate the glycosylation profile and concentration of a subsequent sample. We believe this could pave a new way for SPR to be used in product quality monitoring.
Our analysis framework could be applied to study heterogeneity in the analyte solution hailing from factors other than glycosylation, such as differences in size or folding, for example.
We propose an algorithm that can estimate the concentration, as long as the dilution factors used to create the dilution series are known (if more than one concentration was injected). However, knowing the concentration still leads to more precise composition estimates, as errors in the composition estimate can be compensated for by errors in the concentration estimate, and vice versa. The active concentration can be found by using a surface with a high density of immobilized ligand and a low flow rate, so that the binding process is limited by mass transport limitations36,47,48,49.
The system is assumed to be limited by the binding kinetics in both of our algorithms, rather than mass transport. If mass transport limitations are only moderate, the system should behave similarly to a kinetically limited system, except with slower kinetics. There should not be an effect on the binding affinity nor the equilibrium behavior. The algorithms presented here might be usable on systems moderately limited by mass transport limitations (as long as equilibrium may be reached), but the kinetics uncovered would be biased11,48,50.
We extended the previously existing framework for the characterization of analyte mixtures via SPR biosensing. This was achieved by taking advantage of the ability of SPR biosensors to perform experiments at different temperatures. The developed algorithms extracted analyte-specific kinetics from mixtures of analytes and then used these kinetic parameters to identify the composition of other unknown mixtures. The algorithms were found to be precise and robust. Performing experiments at multiple temperatures effectively reduced the number of mixtures that was required to extract analyte-specific kinetics. A couple limitations remain in our analysis framework. First, a simple 1:1 binding model is assumed, which might not be appropriate for all systems. Second, the identification of the kinetics still requires prior knowledge of the composition and the concentration of the mixtures used in our first algorithm, which may be obtained with orthogonal characterization techniques.
We believe that mixture analysis opens a new avenue for the use of SPR both as a valuable tool to facilitate and accelerate the development of bioprocesses and as a monitoring tool for quality control. With more emphasis being put on process analytical tools in recent years16,19, developing new data analysis methods and experimental assays will ensure SPR biosensing remain an asset for research and industry.
The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.
De Crescenzo, G., Boucher, C., Durocher, Y. & Jolicoeur, M. Kinetic characterization by surface plasmon resonance-based biosensors: Principle and emerging trends. Cell. Mol. Bioeng. 1, 204–215. https://doi.org/10.1007/s12195-008-0035-5 (2008).
CAS  Article  Google Scholar 
Guo, X. Surface plasmon resonance based biosensor technique: a review. J. Biophotonics 5, 483–501. https://doi.org/10.1002/jbio.201200015 (2012).
CAS  Article  PubMed  Google Scholar 
Homola, J. Present and future of surface plasmon resonance biosensors. Anal. Bioanal. Chem. 377, 528–539. https://doi.org/10.1007/s00216-003-2101-0 (2003).
CAS  Article  PubMed  Google Scholar 
Couture, M., Zhao, S. S. & Masson, J.-F. Modern surface plasmon resonance for bioanalytics and biophysics. Phys. Chem. Chem. Phys. 15, 11190–11216. https://doi.org/10.1039/C3CP50281C (2013).
CAS  Article  PubMed  Google Scholar 
Prabowo, B. A., Purwidyantri, A. & Liu, K.-C. Surface plasmon resonance optical sensor: A review on light source technology. Biosensors (Basel) 8, 80. https://doi.org/10.3390/bios8030080 (2018).
CAS  Article  Google Scholar 
Myszka, D. G. Improving biosensor analysis. J. Mol. Recognit.: JMR 12, 279–284. https://doi.org/10.1002/(sici)1099-1352(199909/10)12:5%3c279::Aid-jmr473%3e3.0.Co;2-3 (1999).
CAS  Article  PubMed  Google Scholar 
Önell, A. & Andersson, K. Kinetic determinations of molecular interactions using Biacore—Minimum data requirements for efficient experimental design. J. Mol. Recognit. 18, 307–317. https://doi.org/10.1002/jmr.745 (2005).
CAS  Article  PubMed  Google Scholar 
Karlsson, R., Katsamba, P. S., Nordin, H., Pol, E. & Myszka, D. G. Analyzing a kinetic titration series using affinity biosensors. Anal. Biochem. 349, 136–147. https://doi.org/10.1016/j.ab.2005.09.034 (2006).
CAS  Article  PubMed  Google Scholar 
De Crescenzo, G., Woodward, L. & Srinivasan, B. Online optimization of surface plasmon resonance-based biosensor experiments for improved throughput and confidence. J. Mol. Recognit. 21, 256–266. https://doi.org/10.1002/jmr.894 (2008).
CAS  Article  PubMed  Google Scholar 
Mehand, M. S., De Crescenzo, G. & Srinivasan, B. Increasing throughput of surface plasmon resonance-based biosensors by multiple analyte injections. J. Mol. Recognit.: JMR 25, 208–215. https://doi.org/10.1002/jmr.2172 (2012).
Article  PubMed  Google Scholar 
Si Mehand, M., De Crescenzo, G. & Srinivasan, B. On-line kinetic model discrimination for optimized surface plasmon resonance experiments. J. Mol. Recognit.: JMR 27, 276–284. https://doi.org/10.1002/jmr.2358 (2014).
CAS  Article  PubMed  Google Scholar 
Mehand, M. S., Srinivasan, B. & De Crescenzo, G. Optimizing multiple analyte injections in surface plasmon resonance biosensors with analytes having different refractive index increments. Sci. Rep. 5, 15855. https://doi.org/10.1038/srep15855 (2015).
ADS  CAS  Article  PubMed  PubMed Central  Google Scholar 
Zhang, Y., Forssén, P., Fornstedt, T., Gulliksson, M. & Dai, X. An adaptive regularization algorithm for recovering the rate constant distribution from biosensor data. Inverse Problems Sci. Eng. 26, 1464–1489. https://doi.org/10.1080/17415977.2017.1411912 (2018).
MathSciNet  Article  MATH  Google Scholar 
Forssén, P. et al. Reliable strategy for analysis of complex biosensor data. Anal. Chem. 90, 5366–5374. https://doi.org/10.1021/acs.analchem.8b00504 (2018).
CAS  Article  PubMed  PubMed Central  Google Scholar 
Gaudreault, J., Liberelle, B., Durocher, Y., Henry, O. & De Crescenzo, G. Determination of the composition of heterogeneous binder solutions by surface plasmon resonance biosensing. Sci. Rep. 11, 3685. https://doi.org/10.1038/s41598-021-83268-z (2021).
ADS  CAS  Article  PubMed  PubMed Central  Google Scholar 
Simon, L. L. et al. Assessment of Recent Process Analytical Technology (PAT) Trends: A Multiauthor Review. Org. Process Res. Dev. 19, 3–62. https://doi.org/10.1021/op500261y (2015).
CAS  Article  Google Scholar 
Grangeia, H. B., Silva, C., Simões, S. P. & Reis, M. S. Quality by design in pharmaceutical manufacturing: A systematic review of current status, challenges and future perspectives. Eur. J. Pharm. Biopharm. 147, 19–37. https://doi.org/10.1016/j.ejpb.2019.12.007 (2020).
CAS  Article  PubMed  Google Scholar 
Yu, L. X. & Kopcha, M. The future of pharmaceutical quality and the path to get there. Int. J. Pharm. 528, 354–359. https://doi.org/10.1016/j.ijpharm.2017.06.039 (2017).
CAS  Article  PubMed  Google Scholar 
Gaudreault, J., Forest-Nault, C., De Crescenzo, G., Durocher, Y. & Henry, O. On the Use of Surface Plasmon Resonance-Based Biosensors for Advanced Bioprocess Monitoring. Processes 2021, 9. https://doi.org/10.3390/pr9111996 (1996).
CAS  Article  Google Scholar 
Okazaki, A. et al. Fucose depletion from human IgG1 oligosaccharide enhances binding enthalpy and association rate between IgG1 and FcgammaRIIIa. J. Mol. Biol. 336, 1239–1249. https://doi.org/10.1016/j.jmb.2004.01.007 (2004).
CAS  Article  PubMed  Google Scholar 
Kanda, Y. et al. Comparison of biological activity among nonfucosylated therapeutic IgG1 antibodies with three different N-linked Fc oligosaccharides: the high-mannose, hybrid, and complex types. Glycobiology 17, 104–118. https://doi.org/10.1093/glycob/cwl057 (2007).
CAS  Article  PubMed  Google Scholar 
Shibata-Koyama, M. et al. The N-linked oligosaccharide at Fc gamma RIIIa Asn-45: an inhibitory element for high Fc gamma RIIIa binding affinity to IgG glycoforms lacking core fucosylation. Glycobiology 19, 126–134. https://doi.org/10.1093/glycob/cwn110 (2009).
CAS  Article  PubMed  Google Scholar 
Subedi, G.P., & Barb, A.W. The immunoglobulin G1 N-glycan composition affects binding to each low affinity Fc γ receptor. mAbs 2016, 8, 1512–1524, doi:https://doi.org/10.1080/19420862.2016.1218586.
Falconer, D. J., Subedi, G. P., Marcella, A. M. & Barb, A. W. Antibody Fucosylation Lowers the FcγRIIIa/CD16a Affinity by Limiting the Conformations Sampled by the N162-Glycan. ACS Chem Biol 13, 2179–2189. https://doi.org/10.1021/acschembio.8b00342 (2018).
CAS  Article  PubMed  PubMed Central  Google Scholar 
Cambay, F.; Henry, O.; Durocher, Y.; De Crescenzo, G. Impact of N-glycosylation on Fcγ receptor/IgG interactions: unravelling differences with an enhanced surface plasmon resonance biosensor assay based on coiled-coil interactions. mAbs 2019, 11, 435–452, doi:https://doi.org/10.1080/19420862.2019.1581017.
Cambay, F. et al. Glycosylation of Fcγ receptors influences their interaction with various IgG1 glycoforms. Mol. Immunol. 121, 144–158. https://doi.org/10.1016/j.molimm.2020.03.010 (2020).
CAS  Article  PubMed  Google Scholar 
Forest-Nault, C.; Gaudreault, J.; Henry, O.; Durocher, Y.; De Crescenzo, G. On the Use of Surface Plasmon Resonance Biosensing to Understand IgG-FcγR Interactions. Int. J. Mol. Sci. 2021, 22, doi:https://doi.org/10.3390/ijms22126616.
Day, Y. S., Baird, C. L., Rich, R. L. & Myszka, D. G. Direct comparison of binding equilibrium, thermodynamic, and rate constants determined by surface- and solution-based biophysical methods. Prot. Sci. 11, 1017–1025. https://doi.org/10.1110/ps.4330102 (2002).
CAS  Article  Google Scholar 
Navratilova, I. et al. Thermodynamic benchmark study using Biacore technology. Anal. Biochem. 364, 67–77. https://doi.org/10.1016/j.ab.2007.01.031 (2007).
CAS  Article  PubMed  Google Scholar 
Mehand, M. S., Srinivasan, B. & De Crescenzo, G. Estimation of analyte concentration by surface plasmon resonance-based biosensing using parameter identification techniques. Anal. Biochem. 419, 140–144. https://doi.org/10.1016/j.ab.2011.08.051 (2011).
CAS  Article  PubMed  Google Scholar 
Biacore. Biacore T100 Software Handbook Uppsala, Sweden, 2006.
Evans, N. D. et al. Structural identifiability of surface binding reactions involving heterogeneous analyte: Application to surface plasmon resonance experiments. Automatica 49, 48–57. https://doi.org/10.1016/j.automatica.2012.09.015 (2013).
MathSciNet  Article  MATH  Google Scholar 
Supuran, C. T. How many carbonic anhydrase inhibition mechanisms exist?. J. Enzyme Inhib. Med. Chem. 31, 345–360. https://doi.org/10.3109/14756366.2015.1122001 (2016).
CAS  Article  PubMed  Google Scholar 
Chavane, N., Jacquemart, R., Hoemann, C. D., Jolicoeur, M. & De Crescenzo, G. At-line quantification of bioactive antibody in bioreactor by surface plasmon resonance using epitope detection. Anal. Biochem. 378, 158–165. https://doi.org/10.1016/j.ab.2008.04.019 (2008).
CAS  Article  PubMed  Google Scholar 
Jacquemart, R. et al. At-line monitoring of bioreactor protein production by surface plasmon resonance. Biotechnol. Bioeng. 100, 184–188. https://doi.org/10.1002/bit.21725 (2008).
CAS  Article  PubMed  Google Scholar 
Karlsson, R. Biosensor binding data and its applicability to the determination of active concentration. Biophys. Rev. 8, 347–358. https://doi.org/10.1007/s12551-016-0219-5 (2016).
CAS  Article  PubMed  PubMed Central  Google Scholar 
Wieland, F.-G., Hauber, A. L., Rosenblatt, M., Tönsing, C. & Timmer, J. On structural and practical identifiability. Curr. Opin. Syst. Biol. 25, 60–69. https://doi.org/10.1016/j.coisb.2021.03.005 (2021).
CAS  Article  Google Scholar 
Raue, A. et al. Structural and practical identifiability analysis of partially observed dynamical models by exploiting the profile likelihood. Bioinformatics 25, 1923–1929. https://doi.org/10.1093/bioinformatics/btp358 (2009).
CAS  Article  PubMed  Google Scholar 
Davis, T. M. & Wilson, W. D. Determination of the refractive index increments of small molecules for correction of surface plasmon resonance data. Anal. Biochem. 284, 348–353. https://doi.org/10.1006/abio.2000.4726 (2000).
CAS  Article  PubMed  Google Scholar 
Nakajima, H. et al. Kinetic Analysis of Binding between Shiga Toxin and Receptor Glycolipid Gb3Cer by Surface Plasmon Resonance*. J. Biol. Chem. 276, 42915–42922. https://doi.org/10.1074/jbc.M106015200 (2001).
CAS  Article  PubMed  Google Scholar 
Pearson, J. T. et al. Surface Plasmon Resonance Analysis of Antifungal Azoles Binding to CYP3A4 with Kinetic Resolution of Multiple Binding Orientations. Biochemistry 45, 6341–6353. https://doi.org/10.1021/bi0600042 (2006).
CAS  Article  PubMed  Google Scholar 
Perlmann, G. E. & Longsworth, L. G. The Specific Refractive Increment of Some Purified Proteins. J. Am. Chem. Soc. 70, 2719–2724. https://doi.org/10.1021/ja01188a027 (1948).
CAS  Article  PubMed  Google Scholar 
Zhao, H., Brown, P. H. & Schuck, P. On the distribution of protein refractive index increments. Biophys J 100, 2309–2317. https://doi.org/10.1016/j.bpj.2011.03.004 (2011).
CAS  Article  PubMed  PubMed Central  Google Scholar 
Tan, C.-Y. & Huang, Y.-X. Dependence of Refractive Index on Concentration and Temperature in Electrolyte Solution, Polar Solution, Nonpolar Solution, and Protein Solution. J. Chem. Eng. Data 60, 2827–2833. https://doi.org/10.1021/acs.jced.5b00018 (2015).
CAS  Article  Google Scholar 
Hossler, P., Khattak, S. F. & Li, Z. J. Optimal and consistent protein glycosylation in mammalian cell culture. Glycobiology 19, 936–949. https://doi.org/10.1093/glycob/cwp079 (2009).
CAS  Article  PubMed  Google Scholar 
Yamane-Ohnuki, N. et al. Establishment of FUT8 knockout Chinese hamster ovary cells: An ideal host cell line for producing completely defucosylated antibodies with enhanced antibody-dependent cellular cytotoxicity. Biotechnol. Bioeng. 87, 614–622. https://doi.org/10.1002/bit.20151 (2004).
CAS  Article  PubMed  Google Scholar 
Pol, E. et al. Evaluation of calibration-free concentration analysis provided by Biacore™ systems. Anal Biochem 510, 88–97. https://doi.org/10.1016/j.ab.2016.07.009 (2016).
CAS  Article  PubMed  Google Scholar 
Goldstein, B., Coombs, D., He, X., Pineda, A. R. & Wofsy, C. The influence of transport on the kinetics of binding to surface receptors: application to cells and BIAcore. J. Mol. Recognit.: JMR 12, 293–299. https://doi.org/10.1002/(sici)1099-1352(199909/10)12:5%3c293::Aid-jmr472%3e3.0.Co;2-m (1999).
CAS  Article  PubMed  Google Scholar 
Christensen, L. L. Theoretical analysis of protein concentration determination using biosensor technology under conditions of partial mass transport limitation. Anal. Biochem. 249, 153–164. https://doi.org/10.1006/abio.1997.2182 (1997).
CAS  Article  PubMed  Google Scholar 
Mason, T., Pineda, A. R., Wofsy, C. & Goldstein, B. Effective rate models for the analysis of transport-dependent biosensor data. Math Biosci 159, 123–144. https://doi.org/10.1016/s0025-5564(99)00023-1 (1999).
CAS  Article  PubMed  MATH  Google Scholar 
Download references
This work was supported by the Natural Sciences and Engineering Research Council of Canada (stipends allocated to JG including one from the PrEEmiuM CREATE program). This work was supported by the TransMedTech Institute (NanoBio Technology Platform) and its main funding partner, the Canada First Research Excellence Fund. The authors would like to thank Romane Oliverio, Benjamin Serafin, Benoît Liberelle and Catherine Forest-Nault for fruitful discussions and revisions during the writing of the manuscript.
These authors contributed equally: Olivier Henry and Gregory De Crescenzo.
Department of Chemical Engineering, Polytechnique Montréal, Centre-Ville Station, P.O. Box 6079, Montreal, QC, H3C 3A7, Canada
Jimmy Gaudreault, Olivier Henry & Gregory De Crescenzo
Life Sciences, NRC Human Health Therapeutics Portfolio, Building Montreal-Royalmount, National Research Council Canada, Montreal, QC, H4P 2R2, Canada
Yves Durocher
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
J.G. performed all experiments and data analysis, Y.D. provided mentorship, O.H. and G.D.C contributed equally to the supervision of this work. All authors reviewed the manuscript.
Correspondence to Olivier Henry or Gregory De Crescenzo.
The authors declare no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Reprints and Permissions
Gaudreault, J., Durocher, Y., Henry, O. et al. Multi-temperature experiments to ease analysis of heterogeneous binder solutions by surface plasmon resonance biosensing. Sci Rep 12, 14401 (2022). https://doi.org/10.1038/s41598-022-18450-y
Download citation
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-022-18450-y
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
Advertisement
Advanced search
Scientific Reports (Sci Rep) ISSN 2045-2322 (online)
© 2022 Springer Nature Limited
Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

source

Hotness Alert: American Singer Madonna Stealing The Glamour Dressed In A Stunning Green Dress – IWMBuzz

Madonna is a marvellous American singer, songwriter, and actress. She is referred as the “Queen of Pop” and is noted for her continual reinvention and versatility in music production, songwriting, and visual presentation. 
The singer has a very unique fashion sense and has to ability to set out trends. Recently, Madonna took to her instagram and shared a string of pictures, looked buzz-worthy.
In the pictures, Madonna is dressed in a gorgeous onepiece dress, that featured green huge leaves all over it, except one side of her breasts. The piece was a unique from the fashion brand Dolce and Gabbana. She also paired her look with black netted gloves, giving her ensemble a bold look.
Madonna paired her dress with several jewellery items, that included statement necklaces, rings and “Provoke” bracelet. She also carried a Versace hand bag. In all the images, Madonna looked breathtaking.
Lastly, Madonna added a picture with an artist, who had her face covered with a mask. She seemed to be the same girlfriend who Madonna was seen kissing on her 64th birthday.
Talking about her 64th birthday, the singer had a blast celebrating her born day in Silicy. As to caption the recent post, Madonna wrote, “Ciao Silicy, you were so good to me”.










Let us know your views in the comments section below and for more updates, stay tuned to IWMBuzz.com
Also Read: Selena Gomez’s Latest Photos Are Worth Watching, See Pictures
Copyright 2022 IWMBuzz.com. All rights reserved

source

Class in session – C-VILLE Weekly

Emily Waters begins her first year as director of the Charlottesville High School orchestra. Photo: John Robinson.
Support C-VILLE Weekly
C-VILLE Weekly is Charlottesville’s leading newspaper. Founded in 1989, it’s been the area’s local source for informative (and informed) stories in news, arts, and living for more than 26 years.
Subscribe to our newsletter
Donate
C-VILLE Weekly l August 17, 2022
C-VILLE Weekly l August 10, 2022
C-VILLE Weekly l August 3 2022
C-VILLE Weekly l July 27, 2022
Best of C-VILLE | 2021
434 | Summer 2022
Made In C-VILLE | Summer2022
Weddings | Spring 2022
Real Estate Weekly | August 17, 2022
Real Estate Weekly | August 10, 2022
Real Estate Weekly | August 3, 2022
Real Estate Weekly | July 27, 2022
Fine Properties | June
Gen Now | July
2022 Power Issue
2022 We Are C-VILLE
News
Culture
Magazines

© Copyright C-ville Weekly

source

Scrum Master vs. Product Owner: What's The Difference? – Forbes

Scrum Master vs. Product Owner: What’s The Difference?  Forbes
source

DJ Software Market Size, Growth 2022 Global Statistics, Industry Trends, Competition Strategies, Revenue Analysis, Key Players, Regional Analysis by Forecast to 2027 with CAGR of 4.79% – Digital Journal

Hi, what are you looking for?
By
Published
DJ Software Market report dissects the main thrusts of the worldwide DJ Software market just as reveals insight into restrictions and limitations. The worldwide market is segmented based on type, application, deals, and area. It contains numbers, regions, revenue, and offers just as the thorough computation of the business chain structure, opportunities, industry news analysis.
Global DJ Software Market research report represents a In-Depth overview of the current market situation and forecast till 2027. The study perhaps a perfect mixture of qualitative and quantitative information highlighting key market developments, challenges, competition industry analysis and new opportunities available and trend within the DJ Software Market. Further, this report gives DJ Software Market size, recent trends, growth, share, development status, market dynamics, cost structure, and competitive landscape. The research report also includes the present market and its growth potentials in the given period of forecast. An exhaustive and professional study of the global DJ Software market report has been completed by industry professionals and presented in the most particular manner to present only the details that matter the most. The report mainly focuses on the most dynamic information of the global market.
Get a sample PDF of the report –https://www.marketreportsworld.com/enquiry/request-sample/21512516
About DJ Software Market:-
The global DJ Software market size was valued at USD 370.0 million in 2021 and is expected to expand at a CAGR of 4.79% during the forecast period, reaching USD 490.0 million by 2027.
DJ Software is the application that DJs use when they want to use a computer to play their music.DJ software have the following applications, such as personal and commercial. And personal was the most widely used area.
The report combines extensive quantitative analysis and exhaustive qualitative analysis, ranges from a macro overview of the total market size, industry chain, and market dynamics to micro details of segment markets by type, application and region, and, as a result, provides a holistic view of, as well as a deep insight into the DJ Software market covering all its essential aspects.
For the competitive landscape, the report also introduces players in the industry from the perspective of the market share, concentration ratio, etc., and describes the leading companies in detail, with which the readers can get a better idea of their competitors and acquire an in-depth understanding of the competitive situation. Further, mergers and acquisitions, emerging market trends, the impact of COVID-19, and regional conflicts will all be considered.
In a nutshell, this report is a must-read for industry players, investors, researchers, consultants, business strategists, and all those who have any kind of stake or are planning to foray into the market in any manner.
The global DJ Software market size is segmented on the basis of application, end user, and region, with focus on manufacturers in different regions. The study has detailed the analysis of different factors that increase the industries growth. This study also provides the scope of different segments and applications that can potentially influence the industry in the future. Pricing analysis is covered in this report according to each type, manufacturer, regional analysis, price. DJ Software Market Share report provides overview of market value structure, cost drivers, various driving factors and analyze industry atmosphere, then studies global outline of industry size, demand, application, revenue, product, region and segments. In addition, this report introduces market competition situation among the distributers and manufacturers profile, besides, market value analysis and cost chain structure are covered in this report.
Get a Sample Copy of the DJ Software Market Report 2022
TOP Manufactures in DJ Software Market are: –
DJ Software Market Forecast by regions, type and application, with sales and revenue, from 2021 to 2027. DJ Software Market Share, distributors, major suppliers, changing price patterns and the supply chain of raw materials is highlighted in the report.DJ Software Market Size report provides important information regarding the total valuation that this industry holds presently and it also lists the segmentation of the market along with the growth opportunities present across this business vertical.This Report Focuses on the DJ Software Market manufacturers, to study the sales, value, market share and development plans in the future. It is Define, describe and forecast the DJ Software Market Growth by type, application, and region to Study the global and key regions market potential and advantage, opportunity and challenge, restraints and risks. Know significant trends and factors driving or inhibiting the DJ Software Market growth opportunities in the market for stakeholders by identifying the high growth segments. Strategically it examines each submarket with respect to individual growth trend and their contribution to the DJ Software Market.
Enquire before purchasing this report – https://www.marketreportsworld.com/enquiry/pre-order-enquiry/21512516
On the basis of product, this report displays the production, revenue, price, market share and growth rate of each type, primarily split into
On the basis of the end users/applications, this report focuses on the status and outlook for major applications/end users, consumption (sales), market share and growth rate for each application, including
The Global DJ Software Market Trends,development and marketing channels are analysed. Finally, the feasibility of new investment projects is assessed and overall research conclusions offered.The global DJ Software Market Growth is anticipated to rise at a considerable rate during the forecast period, between 2021 and 2027. In 2021, the market was growing at a steady rate and with the rising adoption of strategies by key players, the market is expected to rise over the projected horizon.
To Understand How COVID-19 Impact is Covered in This Report. Request Sample copy of the report at – : https://www.marketreportsworld.com/enquiry/request-covid19/21512516
DJ Software Market Trend for Development and marketing channels are analysed. Finally, the feasibility of new investment projects is assessed and overall research conclusions offered. DJ Software Market Report also mentions market share accrued by each product in the DJ Software market, along with the production growth.
Regions are covered in Chapter 5, 6, 7, 8, 9, 10, 13:
North America (Covered in Chapter 6 and 13)
Europe (Covered in Chapter 7 and 13)
Asia-Pacific (Covered in Chapter 8 and 13)
Middle East and Africa (Covered in Chapter 9 and 13)
South America (Covered in Chapter 10 and 13)
Chapters Included in DJ Software Market Report: –
Purchase this report (Price 3250 USD for a single-user license) –https://www.marketreportsworld.com/purchase/21512516
Some of the Key Questions Answered in this Report:
Detailed TOC of Global DJ Software Market Development Strategy Pre and Post COVID-19, by Corporate Strategy Analysis, Landscape, Type, Application, and Leading 20 Countries
1 Market Overview
1.1 Product Definition and Market Characteristics
1.2 Global DJ Software Market Size
1.3 Market Segmentation
1.4 Global Macroeconomic Analysis
1.5 SWOT Analysis
2. Market Dynamics
2.1 Market Drivers
2.2 Market Constraints and Challenges
2.3 Emerging Market Trends
2.4 Impact of COVID-19
2.4.1 Short-term Impact
2.4.2 Long-term Impact
3 Associated Industry Assessment
3.1 Supply Chain Analysis
3.2 Industry Active Participants
3.2.1 Suppliers of Raw Materials
3.2.2 Key Distributors/Retailers
3.3 Alternative Analysis
3.4 The Impact of Covid-19 From the Perspective of Industry Chain
4 Market Competitive Landscape
4.1 Industry Leading Players
4.2 Industry News
4.2.1 Key Product Launch News
4.2.2 MandA and Expansion Plans
5 Analysis of Leading Companies
6 Market Analysis and Forecast, By Product Types
6.1 Global DJ Software Sales, Revenue and Market Share by Types(2017-2022)
6.1.1 Global DJ Software Sales and Market Share by Types(2017-2022)
6.1.2 Global DJ Software Revenue and Market Share by Types (2017-2022)
6.1.3 Global DJ Software Price by Types (2017-2022)
6.2 Global DJ Software Market Forecast by Types (2017-2022)
6.2.1 Global DJ Software Market Forecast Sales and Market Share by Types(2022-2027)
6.2.2 Global DJ Software Market Forecast Revenue and Market Share by Types(2022-2027)
6.3 Global DJ Software Sales, Price and Growth Rate by Types(2017-2022)
7 Market Analysis and Forecast, By Applications
7.1 Global DJ Software Sales, Revenue and Market Share by Applications(2017-2022)
7.1.1 Global DJ Software Sales and Market Share by Applications(2017-2022)
7.1.2 Global DJ Software Revenue and Market Share by Applications(2017-2022)
8 Market Analysis and Forecast, By Regions
8.1 Global DJ Software Sales by Regions(2017-2022)
8.2 Global DJ Software Market Revenue by Regions(2017-2022)
8.3 Global DJ Software Market Forecast by Regions(2022-2027)
9 North America DJ Software Market Analysis
9.1 Market Overview and Prospect Analysis
9.2 North America DJ Software Market Sales and Growth Rate(2017-2022)
9.3 North America DJ Software Market Revenue and Growth Rate(2017-2022)
9.4 North America DJ Software Market Forecast
9.5 The Influence of COVID-19 on North America Market
9.6 North America DJ Software Market Analysis by Country
10 Europe DJ Software Market Analysis
10.1 Market Overview and Prospect Analysis
10.2 Europe DJ Software Market Sales and Growth Rate(2017-2022)
10.3 Europe DJ Software Market Revenue and Growth Rate(2017-2022)
10.4 Europe DJ Software Market Forecast
10.5 The Influence of COVID-19 on Europe Market
10.6 Europe DJ Software Market Analysis by Country
11 Asia-Pacific DJ Software Market Analysis
11.1 Market Overview and Prospect Analysis
11.2 Asia-Pacific DJ Software Market Sales and Growth Rate(2017-2022)
11.3 Asia-Pacific DJ Software Market Revenue and Growth Rate(2017-2022)
11.4 Asia-Pacific DJ Software Market Forecast
11.5 The Influence of COVID-19 on Asia Pacific Market
11.6 Asia-Pacific DJ Software Market Analysis by Country
12 South America DJ Software Market Analysis
12.1 Market Overview and Prospect Analysis
12.2 South America DJ Software Market Sales and Growth Rate(2017-2022)
12.3 South America DJ Software Market Revenue and Growth Rate(2017-2022)
12.4 South America DJ Software Market Forecast
12.5 The Influence of COVID-19 on South America Market
12.6 South America DJ Software Market Analysis by Country
13 Middle East and Africa DJ Software Market Analysis
13.1 Market Overview and Prospect Analysis
13.2 Middle East and Africa DJ Software Market Sales and Growth Rate(2017-2022)
13.3 Middle East and Africa DJ Software Market Revenue and Growth Rate(2017-2022)
13.4 Middle East and Africa DJ Software Market Forecast
13.5 The Influence of COVID-19 on Middle East and Africa Market
13.6 Middle East and Africa DJ Software Market Analysis by Country
14 Conclusions and Recommendations
14.1 Key Market Findings and Prospects
14.2 Advice for Investors
15 Appendix
15.1 Methodology
15.2 Research Data Source
……Continued
Browse complete table of contents at –https://www.marketreportsworld.com/TOC/21512516
About Us:
Market Reports Worldis the Credible Source for Gaining the Market Reports that will Provide you with the Lead Your Business Needs. Market is changing rapidly with the ongoing expansion of the industry. Advancement in the technology has provided today’s businesses with multifaceted advantages resulting in daily economic shifts. Thus, it is very important for a company to comprehend the patterns of the market movements in order to strategize better. An efficient strategy offers the companies with a head start in planning and an edge over the competitors.
Contact Us:
Market Reports World
Phone:US : +(1) 424 253 0946
UK : +(44) 203 239 8187
Email:[email protected]
Web:https://www.marketreportsworld.com
Other Reports Here:
Glass Concrete Fiber Market Size 2022 Global Trend, Industry News, Industry Demand, Business Growth, Top Key Players Update, Business Statistics and Research Methodology by Forecast to 2028
Neuroscience Antibodies and Assays Market Size, Share 2022 Global Growth, Trends, Industry Analysis, Statistics, Development Status, Key Players and Forecast to 2028
Digital Shelf Label Market Size 2022 Global Trend, Segmentation, Business Growth, Top Key Players Analysis Industry, Opportunities and Forecast to 2028
Redispersible Polymer Powder Market Size, Growth, Share | Global Companies, Consumption, Drivers, Top Leading Countries, Trends, Forces Analysis, Revenue, Challenges and Global Forecast 2030
End Brushes Market Size, Growth 2022 Industry Demand, Global Trend, Industry News, Business Top Key Players Update, Business Statistics and Research Methodology by Forecast to 2028
Zeolite Market Size, Growth 2022 Size, Growth, Global Trends, Comprehensive Research Study, Development Status, Opportunities, Future Plans, Competitive Landscape and Growth by Forecast 2025
Sodium Tripolyphosphate (STPP) Market Size, Share Global Industry Updates, Leading Players, Future Growth, Business Prospects, Forthcoming Developments and Future Investments by Forecast to 2025
Diatomite (Diatomaceous Earth) Market Size 2022-Global Growth, Trends, Industry Analysis, Key Players and Forecast 2025
Polyvinylidene Fluoride (PVDF) Market Share 2022 Global Comprehensive Research Study, Trends, Development Status, Opportunities, Future Plans, Competitive Landscape and Growth by Forecast 2027
Food Service Equipment Market Size, Growth 2022 | Global Companies, Consumption, Drivers, Top Leading Countries, Trends, Forces Analysis, Revenue, Challenges and Global Forecast 2025
Press Release Distributed by The Express Wire
To view the original version on The Express Wire visit DJ Software Market Size, Growth 2022 Global Statistics, Industry Trends, Competition Strategies, Revenue Analysis, Key Players, Regional Analysis by Forecast to 2027 with CAGR of 4.79%

The big northern hemisphere drought is hitting Europe, the US and China.. Complacency is not an option.
A federal judge in Texas has blocked a Biden administration guidance that required hospitals to provide emergency abortions.
Healthcare providers should look to invest in a long-term, scalable solution, especially as a recession is anticipated.
Bus drivers in London are set to strike over pay for two days this weekend, their union announced on Wednesday.
COPYRIGHT © 1998 – 2022 DIGITAL JOURNAL INC. Digital Journal is not responsible for the content of external sites. Read more about our external linking.

source

Concho Valley Live: Master Gardners Fall Symposium – ConchoValleyHomepage.com

ConchoValleyHomepage.com
Please enter a search term.
That’s What I Like About Fridays
Please enter a search term.
by:
Posted:
Updated:
by:
Posted:
Updated:
SAN ANGELO, Texas – The 11th Annual Master Gardners’ Fall Symposium is set for September 10th from 8 a.m. to 4 p.m. at the Texas A & M Center on N. U.S. 87.
“The Concho Valley Master Gardners’ is a volunteer group that works with us,” Allison Watkins with the County Extension Service shared on Concho Valley Live. “Their mission in life is to promote horticulture education and water conservation throughout the Concho Valley.”
Through the Fall Symposium, participants will get the opportunity to sit in for six-session, focusing on ways to take care of your landscape and garden such as: Planting Trees in Your Home Landscape by Garret Mikulik, Lawn Care 101 by Monte Sims, Solution Driven Landscaping with Tam Honaker, Traveling Companions: Plants and People by Carol Reese, Go Forth and Multiply also with Reese and Floral Design and Decor From Your Garden with Lisa Whittlesey.
According to Watkins, most sessions are being led by locals with the two others coming from College Station and out of state.
Although two session leaders are traveling from other areas, the Master Gardeners’ Fall Symposium focuses on Concho Valley landscapes and gardening.
“As you may know, gardening and landscaping is a challenge,” said Watkins. “All of our presentations and the speakers are all always here for the Concho Valley.”
Tickets for the day-long symposium are on sale for $30 through the Concho Valley Master Gardners’ Square Site, by visiting the Extension Office at 113 W. Beauregard or calling (325) 659-6522.
Copyright 2022 Nexstar Media Inc. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.
Thanks for signing up!
Watch for us in your inbox.
Subscribe Now
Thanks for signing up!
Watch for us in your inbox.
Subscribe Now

source

The next iteration of the Knitting Factory coming to the East Village – EV Grieve

An award-winning news site covering the East Village of NYC
Never been to the Houston location, but the Tribeca one was fabulous. Had great musical adventures there. Williamsburg one was just okay. We went there basically twice a year, for Legendary Pink Dots and for Acid Mothers Temple.
Met James Doohan at the Houston Street location many years ago. He was emceeing an avant garde bagpipe festival there. Very amiable, as long as you didn’t ask questions about Star Trek. Also first saw Ashley MacIsaac that night, looking very young but way cool in Docs and a kilt. Good times and great set-up, with a space to talk downstairs and a space to listen and look upstairs. It’d be great to have them back.
The empty swath of space on Stuyvesant St at 3rd Ave (RIP Panya) could be a possible location.
I remember the Houston St location fondly, so this feels like it's coming back full circle. Let the location speculation begin!
Good news – more live music spots, less smoke shops!!
Here's hoping it's the venue adjacent to two boots. Also still hoping Zum Schneider makes a triumphant return!
How about the old Pyramid Club space?
The Knit Factory is coming back??…….Oh,no this can't be. Must be fake news. Doesn't everybody know the East Village, in fact all of Downtown, has been DEAD for years and years??…….It a dirty, filthy, scummy mess overrun with Bros, Whoo-hoo girls and GIANT Rats!!!…..Besides which the grumps and old farts are going be apoplectic when they get wind of this !!……….and if true is this really needed??…..LOUD MUSIC, young punks and dancing girls drinking, smoking weed, and having FUN !?…….Who is responsible for this OUTRAGE??
Wasn't that said before that Pyramid was the target?
Hard to believe rent would be cheaper in Manhattan but maybe this is a sign the real estate crash has begun if rock n roll can return to Manhattan.
Ummmm the Pyramid is in the Pyramid space.
I can verify that the Pyramid is the candidate, ummmm.
Nice unexpected win for the neighborhood!
Huge get if true
Yes!!! This is so great. I wonder if they’ll take over Sidewalk’s old space?
Zum Schneider is sadly likely never returning to the LES, if it does at all. They are however hosting their Oktoberfest at 3 Dollar Bill’s giant yard in Bushwick 9/30-10/9
Post a Comment
 Subscribe in a reader

source

New quantum physics, solving puzzles of Wheeler's delayed choice and a particle's passing N slits simultaneously and quantum oscillator in experiments | Scientific Reports – Nature.com

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Advertisement
Scientific Reports volume 12, Article number: 14410 (2022)
Metrics details
This paper discovers new quantum physics, and gives solutions to puzzles of Wheeler’s delayed choice and a particle’s passing many slits simultaneously by exact quantum physics expressions. We further show new quantum control, new quantum oscillation, new quantum control experiments and new quantum oscillator being able to be installed in quantum communication network etc. We discover that the ability of a photon to hit electrons out in photoelectric effect is complementarily equivalent to the ability of wave of a photon to simultaneously pass through many slits in wave-particle duality. Objective criterion for distinguishing classical and quantum particles is found, and this paper gives applicable realm of quantum theories and new quantum physics expressions of wave-particle duality. All these studies above should be classified as classical and quantum particles, then classical particle and quantum particle wave cannot and can pass many slits, respectively. This paper discovers wave-particle duality’s origin of displaying both wave property from plane wave part of the general Fourier expansion and particle property from the general Fourier expansion coefficients with the particle’s global property and spins etc. We give the superposition state representation of wave-particle duality, further find the collapse of the duality superposition state to wave or particle state. The collapsed wave or particle state is related to the measure of wave or particle property. Then, we explain why sometimes it’s a wave or a particle. Our achieved results are truly tested, and we discover new measured attractive state and quantum wave collapse velocity expression.
In modern physics, the double slit experiments show that microscopic particles can display both classical particle and wave properties, and they show the probabilistic nature of phenomena of quantum mechanics1.
In 1801, the double slit experiment was first finished with light by T. Young2, after a long time, electrons display the same behavior in 19272.
A coherent light source, e.g., a laser beam, lights up one plate pierced by two parallel slits, behind the plate, light of passing through the slits is observed on a screen2,3. The experiments demonstrate that the particle doesn’t form interference pattern when one detects which slit the particle passes through; the particle forms the interference pattern when one doesn’t detect which slit the particle passes through. These results show the principle of wave–particle duality4,5.
Feynman said that all of quantum mechanics can be gleaned from carefully thinking through the implications of this single experiment6, and he presented that if detectors were placed before each slit, the interference pattern would disappear6.
References7,8 used a single electron and biprism to show that each electron interferes with itself as predicted by quantum theory in 1974. Readers of Physics World voted the -electron version of the experiment as the most beautiful experiment9.
The double-slit experiment with electrons and real slits is eventually performed by the original scheme from Feynman, they emit electrons onto the double slits, and via collecting the transmitted electrons with a -electron detector, the double-slit interference pattern is shown10.
For antimatter, Ref.11 demonstrated a single particle interference. Photons, electrons, atoms and even some molecules are measured as a single pulse, the waves of these particles describe the probabilities of absorbing the particles at a specific place on the screen12,13,14.
The which-way experiment displays the complementarity principle that photons can show as either particles or waves, but cannot be observed as both at the same time15. In the history of quantum mechanics, despite the importance of the thought experiment, until the 1970s, technically feasible realizations of the thought experiment were presented16. In order to illustrate various aspects of complementarity, multiple experiments have been performed17.
In 1987, a performed experiment18,19 produced the results that the information could be gotten regarding a particle’s going which path.
Wave–particle duality of C60 molecules was investigated20, and Plasmon-Assisted Two-Slit Transmission was explored in Young’s Experiment21.
Wheeler’s delayed choice experiments showed that after a particle passes through the slits, extracting “which path” information at the slits can seem to retroactively alter its previous behavior22.
The new quantum universe is illustrated in Ref.23. The discovery that particles are discrete packets of energy with wave-like properties resulted in quantum mechanics dealing with atomic and subatomic systems. Furthermore, unification theory of classical statistical uncertainty relation and quantum uncertainty relation and its applications are shown24.
Although quantum theory has achieved the great successes in many aspects, there still is the hard puzzle of the single particle’s passing through Young’s double slits simultaneously, the hard puzzle has been plaguing us up to now for decades, this paper wants to give the solution to the puzzle.
For example: according to current quantum mechanics, any material can have wave-particle duality, therefore, the physical phenomenon in Fig. 1 could exist. But in real world, we don’t see such a physical phenomenon, which is an interesting example of the fundamental puzzle in the theory of quantum mechanics. This paper wants to solve this kind of puzzle.
Coming from Charles Addams, The New Yorker, 1940.
The arrangements of this paper are: Section “Studies on the puzzle of a particle’s passing through many slits simultaneously” gives studies on the puzzle of a particle’s passing through Young’s Double Slits simultaneously; Section “Solutions to Wheeler’s delayed choice puzzle and puzzle of a particle’s passing double slits simultaneously by the physics processes of the exact quantum physics expressions” gives solutions to Wheeler’s delayed choice puzzle and puzzle of a particle’s passing double slits simultaneously by the physics processes of the exact quantum mathematical expressions; Section “New quantum control, quantum oscillation and their experiments” shows new quantum control, new quantum oscillation and their experiments; Section “Discussion, summary and conclusion” gives discussion, summary and conclusion.
In photoelectric effect, light waves cannot knock electrons out; and in a photon’s passing through many slit experiment, a photon cannot pass through many slits at the same time. Namely, the two physical processes, respectively, reflect one aspect of wave-particle duality of quantum particle. On the other hand, in photoelectric effect, photons can knock electrons out; in the many slit experiment, a photon light wave can pass through many slits at the same time. The two physical processes then are complementarily equivalent in wave-particle duality of quantum particle. That is, in wave-particle duality of quantum particle, the first and the second cases use the particle property and the wave property, respectively. Namely, a photon can show as either particle or wave, but cannot be observed as both at the same time for a physics process.
We now generally show them by exact deduction.
In 4-dimensional momentum representation of quantum theory, when considering wave function (phi (vec{p},E)) of momentum representation, one has25
Equation (1) is a general Fourier transformation of ( , phi (vec{p},E)) (about the plane wave energy E and momentum (vec{p})) from the four-dimensional momentum representation state vector ( , phi (vec{p},E)) to the projection of the plane wave basic vector (e^{{i(vec{p} cdot vec{r} – tE)/hbar }}) and making integration for getting ( , psi (vec{r},t)), which make ( , psi (vec{r},t)) have not only the characteristics of the probabilistic state vector of the particle but also the characteristics of the plane wave, i.e., make ( , psi (vec{r},t)) have the state vector characteristics of wave-particle duality.
Because the momentum representation state vector ( , phi (vec{p},E)) is nonlocal, it also reflects that the system has the global characteristics of momentum (vec{p}) and energy (E), this global property can be the integrity of the particle, e.g., even including different physics qualities, e.g., spin, since the different qualities are not related to space coordinates.
Therefore, the expression (1) exactly shows wave-particle duality’s origin which displays that the wave property is originating from the plane wave part of the general Fourier expansion, and the particle property is originating from the general Fourier expansion coefficients with the particle’s global property even including different spins.
Therefore, we discover, for arbitrary particle, on an aspect, it propagates with the plane wave of the four-dimensional momentum ((vec{p},E)) as the propagation amplitude of the plane wave; on another aspect, it moves as a particle with the four-dimensional momentum ((vec{p},E)). Especially, when the expanding coefficients have different spins, it moves as a particle with both the four-dimensional momentum ((vec{p},E)) and the different spins, which are the new true physics and the new physical pictures, and uncover the corresponding expressions’ contributions of both wave part and particle part of wave-particle duality origin. Namely, Eq. (1) is the function of unified expression of wave-particle duality.
A little bit of philosophical insight on what this work means that the unified expression of wave-particle duality is just the superposition state of wave-particle duality, and the superposition state of wave-particle duality is physically real.
Furthermore, the infinite big momenta and energy show their corresponding to infinite big velocity in Eq. (1), and then the infinite big velocity is included, i.e., the wave function (1) of coordinate representation has the contribution of infinite big momentum or speed, namely, the wave function at any spatial and time points has the contributions from negative to positive infinite big momenta or speeds. Similarly, when we do the inverse Fourier transformation of Eq. (1) about whole spacetime coordinates, we find that the wave function of 4-dimentional momentum representation has the contributions of the whole 4-dimentional spacetime, i.e., the wave function at any 4-dimentialal momentum spatial point has the contributions from the whole spacetime. Thus, the above both cases just the reasons that Feynman path integral can be done in whole 4-dimentional spacetime or momentum space.
Using Eq. (1), we have wave function of momentum representation at time t
On the other hand, using Huygens’ Principle, one has the basic wave analysis:
Every point of a wave front may be considered the source of secondary wavelets that spread out in all directions with a speed equal to the speed of propagation of the waves. What this means is that when one has a wave, he can view the “edge” of the wave as actually creating a series of circular waves. These waves combine together in most cases to just continue the propagation, and in some cases there are significant observable effects. The wave front can be viewed as the line tangent to all of these circular waves26.
Further using Eq. (1) and Huygens principle above, we have N subwave functions through N slits
where j = 1,2,……,N. No losing generality and for simplicity, taking N = 2 just shows the up slit and down slit, respectively, in Young’s Double Slits in Fig. 2.
Interference of a particle plane wave in Young’s double slit experiment.
Therefore, Eqs. (1)–(3) can also be seen as a kind of expressions of Huygens principle. Consequently, these Fourier expansions physically imply new physics, and are not only just the mathematical tools.
The superposition density function of two subwaves is just Eq. (5) in Section “Solutions to Wheeler’s delayed choice puzzle and puzzle of a particle’s passing double slits simultaneously by the physics processes of the exact quantum physics expressions”, the interference terms of the two subwaves in Fig. 2 are just the third term and fourth term in Eq. (5).
These properties are exactly conforming to the plane wave properties of the single particle, thus a particle plane wave can simultaneously pass through N slits, for simplicity, Young’s Double Slits in Fig. 2, Eq. (3) just generally give the both subwave functions that simultaneously pass through N slits, for simplicity, two slits s1 and s2 in Young’s Double Slits, respectively.
The N subwave functions have the same amplitude (phi (vec{p},E)) for some certain (vec{p},E), (e^{{i(vec{p} cdot vec{r}_{j} – tE)/hbar }}) (j = 1, 2,……, N) are just N plane subwave functions in Eq. (3), and the N probabilistic wave functions in Eq. (3) integrate for ((vec{p},E)) from negative infinite to positive infinite, i.e., having considered all possibility, which make the N expressions (3) exact.
The global property of a particle does not allow the single particle to simultaneously pass through N slits, for simplicity, Young’s double slits, in reality, the interference of a particle wave is observed, which just show a particle wave simultaneously does pass the N slits, for simplicity, the double slits, but all theories up to now cannot solve the hard puzzle of a particle’s passing the N slits, e.g., Young’s double slits simultaneously.
The two subwaves (taking N = 2) from Huygens’ Principle (see section “Studies on the puzzle of a particle’s passing through many slits simultaneously”) of a quantum Boson’s or Fermion’s Eq. (3) are entanglement each other, they may form interference of the subwaves, see Fig. 2, their entanglement wave function is
Using Eq. (4), we get the entanglement probabilistic density
Equation (5) means that the single particle wave (1) passes through the double slits in Young’s double slit experiment at time t, which are shown as the two subwaves that are entanglement subwaves (4).
When one measures that the single particle passes slit s1 and at time t1, then he must get (psi (vec{r}_{2} ,t_{1} ) = 0) because of Eq. (4), consequently, using Eq. (5), we deduce
Namely, there is no interference term in Eq. (6), which means that the interference disappears, i.e., the entanglement wave function density (5) is instantly collapsing into Eq. (6), which means that the single particle finally doesn’t simultaneously pass through Young’s Double Slits by the disentanglement collapse in quantum mechanics.
When one measures that the single particle wave passes slit s2, the studies are similar.
When one doesn’t measure which way slit that the single particle wave passes, then Eq. (5) shows that there are interference terms, which means the single particle wave (1) simultaneously passes through Young’s Double Slits by the probabilistic entanglement in quantum theory, which is shown as the two subwaves by Eqs. (4) and (5).
Wheeler’s delayed choice puzzle: after a particle passes through the slits, extracting “which path” information at the slit in Young’s double slits can seem to retroactively alter its previous behavior, referencing to22,27,28,29 and its references.
Therefore, we find both entanglement and disentanglement for Wheeler’s delayed choice experiment are quantum probabilistic entanglement and disentanglement, e.g., from Eqs. (4)–(6), further show that the quantum probabilistic disentanglement retroactively alters its previous passing through the slits’ behaviors from Eq. (5) with (psi (vec{r}_{2} ,t) ne 0) collapse to Eq. (6) with (psi (vec{r}_{2} ,t_{1} ) = 0) in Wheeler’s delayed choice experiment with (t_{1} > t). Therefore, this paper solves Wheeler’s delayed choice puzzle by the physics process of the exact quantum physics expressions (4)–(6).
We must point out that the distribution of the wave function (psi (vec{r}_{2} ,t) ne 0) is the distribution of the whole space, and this whole space automatically includes the space that the subwave returns from the original path s2 and enters through the other gap s1. Normally, the probability of taking this path is small, but since the particle state is measured to have passed through s1, which makes the particle state (psi (vec{r}_{{1}} ,t_{{1}} )) equivalent to an attractor state, the probability of (psi (vec{r}_{2} ,t) ne 0) and all the other probabilities change to satisfy the physical reality of the particle having passed through s1‘s gap at time t1.
This is the self-consistent requirement of probability conservation. Because the collapse of the wave function (4) is the collapse from all possible states to a measured state with density (6), and the measured state from the collapsed states is equivalent to the measured attractor state with density (6). So, we call the measured state through s1 as the measured attractor state with density (6). This collapse process is in fact the collapse process of all possible states towards the measured attractor state (or simply called as measured attractive state) with density (6).
These are the new physics of the collapse of all possible states (including, naturally, state (psi (vec{r}_{2} ,t) ne 0) that return backwards through the state of s2 and merge into the state of s1) to the measured attractor state (psi (vec{r}_{{1}} ,t_{{1}} )). Consequently, we achieve the new key useful technology for changing the state evolution paths of quantum systems.
Because of using the usual entanglement theory, when the wave function of the system of the two subwaves (psi_{1} (vec{r}_{1} ,t)) and (psi_{2} (vec{r}_{2} ,t)) is expressed as
The system has no entanglement27; for the other cases, there are entanglements27. For example, Eq. (4) has entanglement from quantum probabilistic conservation.
Consequently, this paper finds the quantum probabilistic entanglement and disentanglement from Eqs. (4)–(6).
Therefore, the puzzle of Wheeler’s delayed choice is solved by the physics processes of the probabilistic entanglement and disentanglement expressions (4)–(6). Namely, after a particle wave passes through the slits, extracting “which path” information at one slit in Young’s double slits can seem to retroactively alter its previous behavior, which just originates from quantum disentanglement collapse that (psi (vec{r}_{2} ,t) ne 0) passed through the slit s2 instantly collapses to (psi (vec{r}_{2} ,t_{1} ) = 0) and is incorporated into (rho (vec{r}_{1} ,vec{r}_{2} ,t_{1} ) = rho_{1} (vec{r}_{1} ,t_{1} ) = left| {c_{1} psi (vec{r}_{1} ,t_{1} )} right|^{2}), (intlimits_{{}}^{{}} {} rho_{1} (vec{r}_{1} ,t_{1} )dvec{r}_{1} = 1). Because of having measured that the single particle has passed through the slit s1 at time t1 (> t, which is just the delayed choice’s measuring), that is, this quantum system collapses to the measured state when the system is measured at time t1.
In the same time, the puzzle of the single particle’s simultaneously passing through Young’s Double Slits is also solved by the entanglement wave property (4) and entanglement wave function density (5) of wave-particle duality in quantum theory, namely, by the physics process of the exact quantum physics expressions (1), (4) and (5) in quantum probabilistic entanglement theory without the delayed choice measure.
For the quantum delayed choice experiment studied in the previous section, see Fig. 3, the left side of Fig. 3 is a single photon’s radiating gun that radiates a photon at every interval time T, two detectors further are mounted symmetrically on the up side and down side between Young’s double slits and the screen, i.e., red line detector’s screen M1 is on the up side, red line detector’s screen M2 is on the down side. When the two detectors are closed, there are interference, see Fig. 3.
Interference of photon wave in the new quantum oscillation experiment.
When M1 measures the particle passing through the up S1 slit, M1 is automatically closed (i.e., cancelling the measure, the system is restored to a new state), and M2 is automatically opened after time t (i.e., the system collapses again through measurement). When M2 detects a particle passing through the down slit S2, M2 is closed (to restore the system to a new state), and after time t opens M1 to measure (to cause the system to collapse again).
According to Wheeler’s delayed choice law and the collapse principle of measurement studied in the previous section, by adjusting the symmetric positions of M1 and M2 relative to the double slits, the interval time T and the time parameter t, the oscillation of the particle between paths M1 and M2 must appear, because the theory and the experiments are exactly based on the having been experimentally exactly tested bases of a measured state’s collapse and the returning to a new state after cancelling the measuring in quantum theory, i.e., the bases of the ingenious combinations of two Wheeler’s delayed choice experiments1,2,6,23. Through the new quantum control, it may modulate the new quantum oscillation of the particle by the generalized delayed choice experiment.
In particular, the interval time parameter t has very important physical implications. When the interval time t is equal to zero, i.e., one detector is turned off and the other detector is turned on at the same time. When the time t is not equal to zero, it reflects the degree of response of the quantum system, which is an important response parameter for quantum mechanics to describe the system. The size of the parameter t directly reflects the time effect of the physics reaction of the system. Based on the measured instantaneous collapse effect, this time response parameter should be very, very small, close to zero. We can automatically control the interval time T, the switch of the two detectors and the delayed detection time t of the two detectors by the program design of the computer connecting the two detectors and the single photon’s radiating gun.
For S2 measured by M1 and S1 measured by M2, the above method is equivalent.
For lots of the generalized delayed choice experiments about different bosons and fermions similar to Young’s double slits above can be done. Especially, a similar delayed choice experiment for separating a beam light into two entangled beam lights through a biprism with properties of semi-permeable mirror and semi-reflex mirror can be done very similarly, which is written in the following paper because of the length constraint of this paper.
For a photon’s delayed choice experiment, when the new quantum control and the new quantum oscillation of above the research are realized very well, we are unable to see the photon signal on the screen because of the new quantum oscillation between the oscillation paths.
In this experiment, we first take the double slits away and close the two detectors, we can see the photon signal on the screen, which just show the photon’s particle property; and then we take the double slits backward the original position in Fig. 3 and close the two detectors M1 and M2, we can see the interference of photon wave on the screen, which just show the photon’s wave property; finally, all these investigations above mean that by adjusting the interval time T, the response time parameter t and the symmetrical positions of the two detectors M1 and M2 related to the double slits, so that we cannot see any photon signal on the screen in the new quantum oscillation experiment, then the new quantum control and the new quantum oscillation between the oscillation paths are realized very well.
Therefore, we get the objective criteria that the new quantum control and the new quantum oscillation between the oscillation paths are realized very well, and achieve the new controlling quantum particle’s oscillator being able to be installed in quantum communication network as the usual network and computer need the oscillator to work. All these are useful in quantum theory, quantum computer, quantum communications and so on. Especially, anyone working in this field can repeat the new quantum control and quantum oscillation experiments.
Utilizing the measured response parameter t, we are able to achieve the measured collapse velocity for the quantum system
in which because detector M1 and detector M2 are symmetrically arranged about the double slits, e.g., see red lines in Fig. 3, then (L_{1} = L_{2}) (are the distances between the two measure detectors and the double slits), (Delta) is the distance between the two slits. When time response parameter t (to) 0, (V_{collapse} to infty). The collapse velocity, in the past, are unable to objectively be deduced.
Thus, this paper gives new quantum control, new quantum oscillation, their new realizable very important experiments and the new controlling quantum particle’s oscillator being able to be installed in quantum communication network and quantum computer from the new key useful technology for changing the state evolution paths of quantum systems in quantum theory, quantum computer and quantum communications.
Expression (1) shows wave-particle duality’s origin of displaying both the wave property originating from the plane wave part of the general Fourier expansion and the particle property originating from the general Fourier expansion coefficients with the particle’s global property and even including different spins etc., i.e., wave property and particle property both are expressed by general Fourier expansion in the same time, these are just the reasons that there exists wave-particle duality.
Thus, by using strict physical expressions, we give a physical representation of the wave- particle duality superposition state (1), and find that when measurements are made with devices that measure wave property or particle property, the wave-particle duality state collapses from the superposition state (1) of wave-particle duality to the manifestation state of wave property or particle property. So, these just explain why sometimes it’s particle and sometimes it’s wave.
The N subwaves of Eq. (3) are entanglement, they may form interference between subwaves, their entanglement wave function is Eq. (4) for N = 2.
When one measures that the single particle passes slit s1 at time (t_{{1}}), thus he must get (psi (vec{r}_{2} ,t_{{1}} ) = 0), because of quantum probabilistic entanglement Eq. (4). Consequently, using Eq. (5), we deduce that there is no interference term in Eq. (6), which means that the interference disappears.
When one doesn’t measure which slit that the single particle wave passes, then Eq. (5) shows that there are interference terms.
Therefore, we find that entanglement and disentanglement for Wheeler’s delayed choice are, respectively, quantum probabilistic entanglement and quantum probabilistic disentanglement from Eqs. (4)–(6). Furthermore, this paper solves Wheeler’s delayed choice puzzle by the physics process of the exact quantum physics expressions (4)–(6).
In the photoelectric effect, one needs to think of light as particles, otherwise the electrons will be unable to be beaten out by light wave. That is, in the wave-particle duality, the particle property is only considered, wave property is not considered, and then the photoelectric effect is able to be explained. Similarly, in passing N slits, no losing generality for N = 2 Young’s double slit experiment, the particle property is only considered, it is impossible for a photon to pass through two slits at the same time. But if we only think about wave property (because at this time, all particle properties are transformed into the particle’s wave properties with particle property as propagating coefficients that are just creation and annihilation operators of particles in quantum field theory), it is perfectly natural for the single photon wave to pass through the double slits at the same time.
In photoelectric effect and Young’s double slit experiments, the single photon particle can hit electrons out, and the single photon wave can go through the N slits at the same time. Therefore, this paper discovers that the two cases are complementarily equivalent (i.e., the ability of the single photon to hit electron out is complementarily equivalent to the ability of wave of the single photon to pass through the N slits, e.g., Young’s double slits at the same time in wave-particle duality, which is quantum physics of general bosonic systems).
Similarly, quantum particles such as electrons, neutrons, protons, photons and so on mean that wave of their single quantum particle can simultaneously pass through the N slits because at this time, all particle properties are transformed into the particle’s wave properties, rather than individual quantum particle may pass through the N slits at the same time, which is the new quantum physics of general Boson or Fermion particle systems.
This indicates that this effect is a quantum effect, because quantum effect has wave-particle duality. When considering wave property, we cannot consider particle property, because the whole quantum particle system is behaving like wave property (Just as in the photoelectric effect, considering the particle nature of the system, people cannot consider wave property, otherwise the photoelectric effect cannot appear). Namely, the single quantum particle can show as either particle or wave, but cannot be observed as the both at the same time at the same place for a physics process.
The single classical particle cannot simultaneously pass through the N slits, while wave of the single quantum particle can pass through the N slits at the same time, these can be viewed as the objective criterion for distinguishing classical and quantum particles, which is the new quantum physics of general particle systems.
Consequently, we discover the objective criterion for distinguishing classical and quantum particles. Thus, the objective criterion not only objectively gives the applicable realm of quantum mechanics that quantum particles satisfy, but also is key and very useful in quantum physics and science, which will greatly help us to distinguish, investigate and understand physics world and science. Therefore, for the single particle to simultaneously pass through the N slit experiment, needing to divide the studies into classical case and quantum case, the single classical particle cannot pass, while the single quantum particle wave may pass, which is the new quantum physics of general particle systems, then all the problems are solved.
For example, using the discovered objective criterion for classical and quantum particles, the puzzle of the example appearing in above Fig. 1 does not exist, which is the new quantum physics of general particle systems. Because as a person he is a macroscopic classical object, there is not such wave property to produce the physical phenomenon in Fig. 1.
The achieved theoretical results in this paper are also achieved experimentally by the experimental analysis and investigations30. By solving the delayed choice puzzle, we have obtained both the new measured attractor state concept and a new practical useful technique for controlling the changing evolutional paths of the wave function.
Therefore, this paper discovers new quantum physics, and gives solutions to Wheeler’s delayed choice puzzle and a particle’s simultaneously passing many slit puzzle by physics processes of exact quantum physics expressions, further shows new quantum control and new quantum oscillation.
This paper shows new very important feasible experiments of new quantum control and new quantum oscillation, and gets the objective criteria realizing the new quantum control and the new quantum oscillation, further achieves the new controlling quantum particle’s oscillator being able to be installed in quantum communication network and quantum computer as the usual network and computer need the oscillator to work. Furthermore, various quantum oscillators can be similarly constructed according to the similar principle31.
All these studies in this paper are very important and useful in quantum theory, quantum computer, quantum communications and so on, and are consistent with all the current physics theories and experiments. Consequently, all current articles, (text)books and concepts relevant to the results of this paper would be rewritten, updated and supplied. Following this paper, a lot of works can be done, which means that this paper would be widely propagated and largely cited.
All data are included in this paper, if there are any other needs, we shall provide all the needs from our emails. Dr. C. Huang should be contacted if someone wants to request the data from this study.
French, A. P. & Taylor, E. F. An Introduction to Quantum Physics. (Norton, 1978). ISBN 978-0-393-09106-9.
Feynman, R. P., Robert, B. L. & Sands, M. The Feynman Lectures on Physics, Vol. 3 (Addison-Wesley, 1965). ISBN 978-0201021189.
Lederman, L. M. & Christopher, T. H. Quantum Physics for Poets, 102–111 (Prometheus Books, 2011). ISBN 978-1616142810.
Feynman, R. P. There is one lucky break however—electrons behave just like light. in The Feynman Lectures on Physics, Vol. 3. (Quantum Mechanics, 1965).
Davisson, C. J. The diffraction of electrons by a crystal of nickel. Bell Syst. Tech. J. 7, 90–105. https://doi.org/10.1002/j.1538-7305.1928.tb00342.x (1928).
Article  Google Scholar 
Feynman, R. P. The Feynman lectures on physics. Quant. Mech. 11, 80–83 (1965).
ADS  Google Scholar 
Merli, P. G., Missiroli, G. F. & Pozzi, G. On the statistical aspect of electron interference phenomena. Am. J. Phys. 44, 306–307 (1976).
ADS  Article  Google Scholar 
Rosa, R. The Merli–Missiroli–Pozzi two-slit electron-interference experiment. Phys. Perspect. 14, 178–194 (2012).
ADS  Article  Google Scholar 
Physics World. The Most Beautiful Experiment (Physics World, 2002).
Frabboni, S. et al. The Young-Feynman two-slits experiment with electrons: Build-up of the interference pattern and arrival-time distribution using a fast-readout pixel detector. Ultramicroscopy 116, 73–76. https://doi.org/10.1016/j.ultramic.2012.03.017 (2012).
CAS  Article  Google Scholar 
Sala, S. et al. First demonstration of antimatter wave interferometry. Sci. Adv. 5, 7610 (2019).
ADS  Article  Google Scholar 
Greene, B. The Fabric of the Cosmos: Space, Time, and the Texture of Reality, 90 (Random House LLC, 2007). ISBN 978-0-307-42853-0.
Donati, O., Missiroli, G. F. & Pozzi, G. An experiment on electron interference. Am. J. Phys. 41, 639–644 (1973).
ADS  Article  Google Scholar 
Nairz, O., Brezger, B., Arndt, M. & Zeilinger, A. Diffraction of complex molecules by structures made of light. Phys. Rev. Lett. 87, 160401 (2001).
ADS  CAS  Article  Google Scholar 
Boscá, D.-P. & María, C. Updating the wave-particle duality. in 15th UK and European Meeting on the Foundations of Physics. (2007). Accessed 21 June 2008.
Bartell, L. Complementarity in the double-slit experiment: On simple realizable systems for observing intermediate particle-wave behavior. Phys. Rev. D. 21, 1698–1699 (1980).
ADS  MathSciNet  Article  Google Scholar 
Zeilinger, A. Experiment and the foundations of quantum physics. Rev. Mod. Phys. 71, S288–S297 (1999).
CAS  Article  Google Scholar 
Mittelstaedt, P., Prieur, A. & Schieder, R. Unsharp particle-wave duality in a photon split-beam experiment. Found. Phys. 17, 891–903 (1987).
ADS  Article  Google Scholar 
Greenberger, D. M. & Yasin, A. Simultaneous wave and particle knowledge in a neutron interferometer. Phys. Lett. A 128, 391–394 (1988).
ADS  CAS  Article  Google Scholar 
Arndt, M. et al. Wave–particle duality of C60 molecules. Nature 401(6754), 680–682. https://doi.org/10.1038/44348 (1999).
ADS  CAS  Article  PubMed  Google Scholar 
Schouten, H. F. et al. Plasmon-assisted two-slit transmission: Young’s experiment revisited. Phys. Rev. Lett. 94, 053901 (2005).
ADS  CAS  Article  Google Scholar 
Jacques, V. et al. Experimental realization of Wheeler’s gedanken experiment. Science 315, 966 (2007).
ADS  CAS  Article  Google Scholar 
Tony, H. The New Quantum Universe (Cambridge University Press, 2003).
Google Scholar 
Sakurai, J. J. & Napolitano, J. Modern Quantum Mechanics 2nd edn. (Addison-Wesley, 2011).
MATH  Google Scholar 
https://www.thoughtco.com/huygens-principle-2699047.
Huang, C. & Huang, Y.-C. Unification theory of classical statistical uncertainty relation and quantum uncertainty relation and its applications. Phys. Lett. A 375, 271 (2011).
ADS  CAS  Article  Google Scholar 
Wilde, M. M. Quantum Information Theory (Cambridge University Press, 2020).
Google Scholar 
Tang, J. S. et al. Realization of quantum Wheeler’s delayed-choice experiment. Nat. Photon. 6, 600 (2012).
ADS  Article  Google Scholar 
Wang, K., Qian, Xu., Zhu, S. & Ma, X.-S. Quantum wave–particle superposition in a delayed-choice experiment. Nat. Photon. 13, 872 (2019).
ADS  CAS  Article  Google Scholar 
Raoch, H. & Werner, S. A. Neutron Interferometry 2nd edn. (Oxford University Press, 2015).
Book  Google Scholar 
Huang, c. Huang, Y.-C. & Nie, Y.-Y. Solutions to puzzles of double Wheeler’s delayed choices and wave collapse velocity and new quantum control, oscillation and oscillator in experiments, submitted to a related journal.
Download references
The work is supported by the U.S. Department of Energy, Contract No. DE-AC02-05CH11231, NSF through Grants PHY-08059, DOE through grant DEFG02-91ER40681 and National Natural Science Foundation of China (No. 11875081).
Institute of Theoretical Physics, Jiangxi Normal University, Nanchang, 330022, China
Changyu Huang, Yong-Chang Huang & Yi-You Nie
Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA, 94720, USA
Changyu Huang
Department of Physics and Astronomy, Purdue University, 525 Northwestern Avenue, W. Lafayette, IN, 47907-2036, USA
Changyu Huang
Institute of Theoretical Physics, Beijing University of Technology, Beijing, 100124, China
Yong-Chang Huang
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
All authours wrote the main manuscript text and contributed to this paper equally. All authors reviewed the manuscript.
Correspondence to Changyu Huang, Yong-Chang Huang or Yi-You Nie.
The authors declare no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Reprints and Permissions
Huang, C., Huang, YC. & Nie, YY. New quantum physics, solving puzzles of Wheeler’s delayed choice and a particle’s passing N slits simultaneously and quantum oscillator in experiments. Sci Rep 12, 14410 (2022). https://doi.org/10.1038/s41598-022-17667-1
Download citation
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-022-17667-1
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
Advertisement
Advanced search
Scientific Reports (Sci Rep) ISSN 2045-2322 (online)
© 2022 Springer Nature Limited
Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

source

Trump legal team asks for neutral 'special master' to review files seized from Mar-a-Lago – WJCL News Savannah

Stay up to speed on all the latest local and national political news.
Video above: Judge says Trump affidavit may be heavily redacted
Lawyers for former President Donald Trump asked a federal judge Monday to halt the FBI’s review of documents recovered from his Florida estate earlier this month until a neutral special master can be appointed to inspect the records.
The request was included in a federal lawsuit, the first filing by Trump’s legal team in the two weeks since the search, that takes broad aim at the FBI investigation into the discovery of classified records at Mar-a-Lago and that foreshadows arguments his lawyers are expected to make as the probe proceeds.
The lawsuit casts the Aug. 8 search, in which the FBI said it recovered 11 sets of classified documents from Mar-a-Lago, as a “shockingly aggressive move.” It also attacks the warrant as overly broad, contends that Trump is entitled to a more detailed description of the records seized from the home and argues that the FBI and Justice Department has long treated him “unfairly.”
“Law enforcement is a shield that protects America. It cannot be used as a weapon for political purposes,” the lawyers wrote Monday. “Therefore, we seek judicial assistance in the aftermath of an unprecedented and unnecessary raid” at Mar-a-Lago.
In a separate statement, Trump said “ALL documents have been previously declassified” — though he has not produced evidence to support that claim — and described the records as having been “illegally seized from my home.” The Justice Department countered in a terse three-sentence statement pointing out that the search had been authorized by a federal judge after the FBI presented probable cause that a crime had been committed.
The filing requests the appointment of a special master not connected the case who would be tasked with inspecting the records recovered from Mar-a-Lago and setting aside those that are covered by executive privilege — a principle that permits presidents to withhold certain communications from public disclosure.
In some other high-profile cases — including investigations involving Rudy Giuliani and Michael Cohen, two of Trump’s personal attorneys — that role has been filled by a former judge.
“This matter has captured the attention of the American public. Merely ‘adequate’ safeguards are not acceptable when the matter at hand involves not only the constitutional rights of President Trump, but also the presumption of executive privilege,” the attorneys wrote.
The lawsuit argues that the records, created during Trump’s White House tenure, are “presumptively privileged.” But the Supreme Court has never determined whether a former president can assert executive privilege over documents, writing in January that the issue is unprecedented and raises “serious and substantial concerns.”
The high court turned down Trump’s plea to block records held by the National Archives from being turned over to the Jan. 6 committee, saying then that his request would have been denied even if he had been the incumbent president, so there was no need to tackle the thorny issue of a former president’s claims.
The lawsuit paints Trump as “fully cooperative” and compliant with investigators, saying members of his personal and household staff were made available for voluntary interviews and quoting him as telling FBI and Justice Department officials during a June visit to Mar-a-Lago, “Whatever you need, just let us know.”
But the chronology of events makes clear that the search took place only after other options to recover classified documents from the home had been incomplete or unsuccessful. In May, for instance, weeks before the search, the Justice Department issued a subpoena for records bearing classification markings.
The Trump team’s lawsuit was assigned to U.S. District Judge Aileen M. Cannon, who was nominated by Trump in 2020 and confirmed by the Senate 56-21 later that year. She is a former assistant U.S. attorney in Florida, handling mainly criminal appeals.
The months-long investigation, which burst into public view with the Mar-a-Lago search, emerged from a referral from the National Archives, which earlier this year retrieved 15 boxes of documents and other items from the estate that should have been turned over to the agency when Trump left the White House. An initial review of that material concluded that Trump had brought presidential records and several other documents that were marked classified to Mar-a-Lago.
FBI and Justice Department officials visited Mar-a-Lago in June and asked to inspect a storage room. Several weeks later, the Justice Department subpoenaed for video footage from surveillance cameras at the estate. After the meeting at Mar-a-Lago, investigators interviewed another witness who told them that there were likely additional classified documents still at the estate, according to a person familiar with the investigation who was not authorized to speak publicly about it.
Separately Monday, a federal judge acknowledged that redactions to an FBI affidavit spelling out the basis for the search might be so extensive as to make the document “meaningless” if released to the public. But he said he continued to believe it should not remain sealed in its entirety because of the “intense” public interest in the investigation.
A written order from U.S. Magistrate Judge Bruce Reinhart largely restates what he said in court last week, when he directed the Justice Department to propose redactions about the information in the affidavit that it wants to remain secret. That submission is due Thursday at noon.
Justice Department officials have sought to keep the entire document sealed, saying disclosing any portion of it risks compromising an ongoing criminal investigation, revealing information about witnesses and divulging investigative techniques. They have advised the judge that the necessary redactions to the affidavit would be so numerous that they would strip the document of any substantive information and make it effectively meaningless for the public.
Reinhart acknowledged that possibility in his Monday order, writing, “I cannot say at this point that partial redactions will be so extensive that they will result in a meaningless disclosure, but I may ultimately reach that conclusion after hearing further from the Government.”
___
Associated Press writers Curt Anderson, Jill Colvin in New York and Mark Sherman contributed to this report.
Hearst Television participates in various affiliate marketing programs, which means we may get paid commissions on editorially chosen products purchased through our links to retailer sites.

source

Las Cruces City Council approves Community of Hope master plan – Las Cruces Sun-News

LAS CRUCES — The city has formally approved a master plan for the future development and expansion of the Mesilla Valley Community of Hope.
The plan, which defines an “ultimate buildout” for the city-owned property at 999 West Amador Avenue, was approved unanimously at the Las Cruces City Council’s regular meeting Aug. 15. 
The city contracted with the Las Cruces-based Huitt-Zollars, Inc. to create the draft plan alongside Architectural Research Consultants, Inc. and Real Estate Advisors. Along with the master plan, an evaluation of the Community of Hope campus’ facilities and recommendations for its management structure were done.
The groups recommended a conceptual site plan, amendments to leases and subleases, an improved maintenance request process, full-time dedicated maintenance staff and new infrastructure and facilities. An assessment conducted by the planning team revealed the Community of Hope campus is badly in need of “major renovations” and capital improvements.
The MVCH is an alliance of nonprofits which assists people who are unhoused or housing insecure, and its campus includes Camp Hope, a transitional community with temporary shelter for the unhoused; Amador Health Center, El Caldito Soup Kitchen, Jardín de Los Niños, a child daycare service; and Casa de Peregrinos, a food bank. 
Amador Health, Jardín, El Caldito and CdP all sublease their spaces from MVCH, which serves as the lead agency. The city-owned land is leased by MVCH.
On June 6, the plan was reviewed and approved by the city’s Development Review Committee, a workgroup consisting of representatives from the Las Cruces Public Works Department, Community Development Department, Las Cruces Utilities, Las Cruces Fire Department, Facilities, Metropolitan Planning Organization, Parks and Recreation Department and Transit.
The master planning group identified $26.1 million in capital improvements that could take place at the campus over the next five years and beyond, including HVAC system improvement, landscaping, erosion control and accessibility for people with disabilities.
“Our goal is to begin to really beautify the campus, which should have been done a long time ago,” said Lorenzo Alba, the executive director of CdP.
More city news:
Michael McDevitt is a city and county government reporter for the Sun-News. He can be reached at 575-202-3205, [email protected] or @MikeMcDTweets on Twitter.

source

Fed to issue long-awaited guidelines for granting master accounts to crypto banks – Cointelegraph

Fed to issue long-awaited guidelines for granting master accounts to crypto banks  Cointelegraph
source

Looking back at summer music releases – The Rice Thresher

By Jacob Pellegrino     8/23/22 9:53pm
Last year, the Thresher published a collection of smaller reviews and impressions of notable album releases from the summer break. One of the big themes of the past year was a slow return to normal, with musicians feeling hopeful about touring again and connecting with fans. This year, many of the pandemic’s effects can still be seen in releases, with many artists leaning into an introspective style after much time spent alone or only with close friends and family. However, many other artists have done exactly the opposite by embracing energy, making music that could be played in the clubs that have recently reopened. From the 29 summer releases I listened to, the following picks convey a range in genres and provide an image of current trends in popular music.
“It’s Almost Dry” by Pusha T
“It’s Almost Dry” is a diptych showing both sides of Pusha T’s art, juxtaposing two unique pockets of ability throughout. In both his solo career and in his hip-hop duo Clipse, Pusha has worked with two producers who rise above the rest at crafting beats that highlight his abilities: Pharrell Williams and Kanye West. In “It’s Almost Dry,” both producers have very different goals that elevate the album and make it a worthy successor to “Daytona.”
Kanye pushed Pusha in the same direction as “Daytona” — hard, sample-heavy “mixtape raps.” Pusha raps over luxurious beats such as the Donny Hathaway sample in “Dreaming of the Past,” an instrumental Kanye originally did not want to part with.
Pharrell went in a completely different direction with Pusha, asking for a “character, that matter-of-fact evil guy,” and encouraging vocal experimentation with more melodic hooks. To get into character, Pusha T went as far as having Joaquin Phoenix’s “Joker” playing on mute while writing bars for the album. On “Brambleton,” Pusha delivers some of his most open lyrics, rapping about his hurt following a former manager’s 2020 interview on VladTV.
Throughout the album, Pusha T never sacrifices his ability to write clever bars that stick with the listener, such as “I been gettin’ at these coins as I’m breakin’ down a brick / Made the jump to each level, Super Mario exists.” In that line, Pusha cleverly overlays his drug dealing past with a classic video game.
One of the classic critiques of Pusha T and similar artists is that they primarily rap about drug dealing. On “It’s Almost Dry,” Pusha again proves how unfounded this critique is. Throughout the album, Pusha T uses drug dealing to convey profound elements of the human condition. As Pusha T has himself noted, his music continues to have power over listeners for the same reason as director Martin Scorcese’s classic gangster movies: they share a similar commitment to quality and innovate with unique stories through the archetype.
“Mr. Morale & The Big Steppers” by Kendrick Lamar
It’d be impossible to write about eye-catching summer music releases and leave out Kendrick Lamar’s “Mr. Morale & The Big Steppers.” His first album since “DAMN.” in 2017 sees Kendrick writing the most personal verses of his career over introspective instrumentals.
Many artists recently, including Saba and Post Malone, have undergone a similar sonic and topical transformation to Kendrick Lamar in the past year. In “Mr. Morale & The Big Steppers,” Kendrick embraces an R&B and soul-influenced sound. At the same time, he also shifts to more melodic vocals that complement the new sound.
However, the shift is not just in the aesthetics. While he’s always created lyrically engaging music, Kendrick embraced a personal level of lyricism hitherto unseen in his work. The best example of this lyrical openness is on “Mother I Sober,” the penultimate track on the album. The track deals with his mother being sexually assaulted in the past and how this traumatic event led her to continuously making sure that he was not hurt in the same way throughout his childhood. Although he was not, the questioning left a mark on Kendrick’s psyche. 
The second half of the song shifts to issues Kendrick himself has dealt with. He begins by telling the listener that he “did it sober sittin’ with [himself]” but “intoxicated [by] a lustful nature that [he] failed to mention.” Lust, not drug abuse, he said, led him to cheat on his longtime partner to distract from his unresolved hurt and insecurity. This addiction and his mother’s experience are linked to the Black community in general as Kendrick laments the “generational curse” caused by sexual abuse. Yet, through the track, he sets himself free from it, embracing therapy and positive change.
Other tracks, like “N95,” call back to the contagious energy of past radio hits like “Humble.” Even at his most commercial on the album, Kendrick still uses the masks worn in the pandemic as a metaphor for the figurative masks people wear to conceal their true selves. However, “We Cry Together” is an Alchemist-produced oddball track that contributes to the theme of domestic issues in families, essentially conveying a domestic argument on tape with dysfunction laid bare for listeners.
Kendrick Lamar turned his lyricism inward throughout “Mr. Morale & The Big Steppers” to a powerful effect that will likely inspire other musicians to be more open in their work.
“Harry’s House” by Harry Styles
During his break from touring during the pandemic, Harry Styles “realized that that home feeling isn’t something that you get from a house; it’s more of an internal thing. You realize that when you stop for a minute.” This reflection is an essential part of “Harry’s House,” an album that refers not to a physical house, but to “a day in [Styles’] mind,” the introspection that happens in his daily life.
On “Cinema,” Styles creates a fun, house-influenced feeling while talking about someone he has feelings for. The chorus emphasizes “I think you’re cool,” but follows it up with the ask, “am I too into you?” While the song sounds confident musically, there is a hesitance to fully commit to the relationship due to personal insecurity rather than the typical superstar fear of being tied down, , which is refreshing to hear from someone as magnetic as Harry Styles. The funky bassline and smooth vocals work well together to create a song worth listening to again.
 The album closes with “Love Of My Life,” a sparsely arranged metaphorical track that seems at first glance to be about someone. However, it is really a tribute to Harry’s home country of England. It’s a fitting end to an album about belonging and an expression of his deep love for where he grew up. It’s a fitting end to an album about belonging and the parts of home beyond the physical location, as it emphasizes the feelings and memories associated with the country rather than simply landmarks
Harry Styles created a personal expression of his everyday musings and thoughts, and the more acoustic stylings of the album mix with influences as diverse as house music to produce something deeply compelling.
“Gemini Rights” by Steve Lacy
Grammy nominated for his production work with The Internet while still a teenager, Steve Lacy has always been an artist to watch. While his previous solo work has always been well made and enjoyable, “Gemini Rights” is a revelation in his artistry. It’s an immensely interesting album, layering a range of styles, instrumentation and vocals to create a deeply rewarding listen.
The first single from the album, “Mercury” is a great introduction to this new expansion to Steve Lacy’s sound, beginning with almost flamenco-like guitar that leads into an immersive sound. The track has incredibly satisfying changes in vocal melody along with a beautiful layered production.
As with much of his previous work, love and relationships are the main theme that pull together the album. “Buttons” is a wistful love song that captures both the initial intensity of falling in love with someone and a reluctance to give up on them after being treated badly. Interestingly, the track consists of one long verse followed by an outro, eschewing traditional song structures. “Give You the World” is a love song that features some of Lacy’s best vocal work on the album. It’s a fitting ending to the album that pulls together the ideas expressed throughout, eschewing traditional song structures. “Give You the World” is a beautiful love song that features some of Lacy’s best vocal work on the album.
Steve Lacy’s artistic evolution on “Gemini Rights” is an incredibly exciting revelation that proves his skill as an artist and his vision for experimenting with new sounds.
“Jack in the Box” by j-hope
As the first BTS member to release an album since the announcement that the members of the group would be focusing on solo music, all eyes were on j-hope. Primarily a rapper, j-hope’s album strays from BTS’ pop releases like “Butter” and “Dynamite” in favor of a darker, hip-hop-influenced sound.
Listeners’ first introduction to this deep dive into j-hope as a solo artist is the song “MORE,” a rumination on his art and dedication to improving. j-hope tells his fans, “my work makes me breathe, so I want more,” emphasizing the necessity of art in his life. The song’s structure is a rapped verse over a traditional hip-hop beat, heavy with snare, that broadens out into a rock chorus backed by a full band. The push and pull between these two elements in the song drives it forward. In the pre-chorus, j-hope’s vocal delivery on “bring it all, I’m doing it all” would be at home on a 90s grungy, alternative rock record.
The brief, 21-minute long album ends with “Arson,” one of the most fun songs vocally on the album. j-hope’s flow has a frantic necessity to it that makes the track interesting and enjoyable to listen to. His placement of stressed syllables every so often creates an almost lurching feel, mirrored in the music video, where he stumbles through a burnt-out landscape with totaled cars. “Arson” uses fire to symbolize rebirth and a fresh start, and now that he’s lit the flame, even he can’t stop it.
“Jack In The Box” is an album that does a great job of differentiating j-hope from the group that he’s been a part of for so many years, leaving the listener wanting to hear more.
“Renaissance” by Beyoncé
After the pandemic kept people inside and away from others, Beyoncé celebrates the reconnection made possible in recent months in “RENAISSANCE.” The project is a vibrant dance album that embraces house production and has an incredible sense of movement.
In an interview with Harper’s Bazaar, Beyoncé said, “With all the isolation and injustice over the past year, I think we are all ready to escape, travel, love and laugh again. I feel a renaissance emerging, and I want to be part of nurturing that escape in any way possible.” While the album is not as personal or lyrically diverse as its predecessor “Lemonade,” it succeeds at its mission, creating music that listeners can move with and gain confidence from.
The lead single from the album, “BREAK MY SOUL,” provides an introduction to Beyoncé’s embrace of house music with sample-heavy house production, a style that continues throughout the album. Although it is a change for Beyoncé, there’s a strong sense of continuity as the music reprises many of her regular themes of self-confidence and strength.
A highlight of the album, “AMERICA HAS A PROBLEM,” is based on “America Has A Problem (Cocaine)” by Kilo Ali. The original 1990 track discusses cocaine and its effects through the metaphor of a woman. In her version, Beyoncé puts herself in the place of the woman in the original track, comparing herself to the addictive nature of the drug. Her comparisons are heard over an infectious beat sampled from the original track. Combined with a catchy refrain, Beyoncé’s characterization of herself as an addictive substance appeals not only to her lover in the track, but also to her listeners.
“RENAISSANCE” is another strong effort from Beyoncé that broadens the musical elements that she incorporates while sticking to her typical subject matter but reflecting it on a new backdrop. It is an album tailor-made for the present, when people are beginning to feel comfortable finding their places again at clubs and other large social gatherings.
A colorful portrait of privilege, egotism and murder emerges in A24’s latest film, “Bodies Bodies Bodies.” Released in theaters Aug. 12, “Bodies” follows the wealthy, assertive Sophie (Amandla Stenberg) as she invites her reserved girlfriend Bee (Maria Bakalova) to meet her friends at a remote mansion. Eerily tranquil shots of attractive 20-somethings floating underwater in the mansion’s pool introduce us to the “Bodies” ensemble: narcissistic David, unassuming Emma, mysterious Jordan, vapid Alice and Alice’s middle-aged, clueless boyfriend Greg. Upon their arrival, Sophie and Bee are met with reluctance and social tensions, which only build as the day progresses. That night, in a blur of drugs and dancing, Sophie suggests a murder mystery game that quickly becomes all too real. While a hurricane rages outside, characters are found bloodily still throughout the house; respective corpses shed light on old grudges, lurid truths and, most importantly, the obnoxious self-absorption of our affluent protagonists. 
On “Nicole,” NIKI looks to her origins to move onwards, collaborating with her younger self to create a work that bridges the gap between two different eras of her life and career. The artist, born Nicole Zefanya, made a splash when she won a nationwide competition to open for Taylor Swift’s Red Tour in Jakarta. At the time, she maintained a YouTube channel, “nzee24,” on which she posted original songs and covers, primarily with a stripped back, guitar-and-vocal-based sound.
Every year from Aug. 1 to Sept. 5., the end of Labor Day weekend, Houston restaurants offer up some of the best deals of the year in the name of charity. Houston Restaurant Weeks — in reality a whole month — is the largest annual fundraiser for the Houston Food Bank, which according to the Restaurant Week website is America’s largest. It includes restaurants from The Woodlands down to the beach in Galveston. Many venues are a short walk or drive from campus, and this fundraiser can be a great way to try out new restaurants at slightly reduced cost. Here are some of the best restaurants participating right outside the hedges.
Please note All comments are eligible for publication by The Rice Thresher.

source

Rapidly predicting Kohn–Sham total energy using data-centric AI | Scientific Reports – Nature.com

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.
Advertisement
Scientific Reports volume 12, Article number: 14403 (2022)
Metrics details
Predicting material properties by solving the Kohn-Sham (KS) equation, which is the basis of modern computational approaches to electronic structures, has provided significant improvements in materials sciences. Despite its contributions, both DFT and DFTB calculations are limited by the number of electrons and atoms that translate into increasingly longer run-times. In this work we introduce a novel, data-centric machine learning framework that is used to rapidly and accurately predicate the KS total energy of anatase ({mathrm{TiO}}_2) nanoparticles (NPs) at different temperatures using only a small amount of theoretical data. The proposed framework that we call co-modeling eliminates the need for experimental data and is general enough to be used over any NPs to determine electronic structure and, consequently, more efficiently study physical and chemical properties. We include a web service to demonstrate the effectiveness of our approach.
Machine Learning (ML) has begun making critical contributions across all the sciences1,2,3,4,5,6. Materials science (MS), a very broad interdisciplinary field that seeks to discover and design new materials, is being particularly impacted7,8,9,10,11 from the theoretical12,13,14,15,16 to the practical—predicting structural properties17,18,19,20,21 and discoverying of new materials e.g., perovskites, nanoparticles, nanoclusters22,23,24,25. A particularly active area has been using ML to get get new types of interatomic potentials—namely the ML potentials26,27,28,29,30,31,32 applied to a wide range of material systems33,34,35,36,37 with accuracies comparable to those of first-principles calculations. Even specializations within ML itself, such as “class imbalances” (where the significance of data is not reflected in its distribution), offer benefits to MS38. The focus of this work is to leverage AI to improve upon the long-standing, traditional quantum modeling technique Density-functional theory (DFT), a computational method developed in 1970s, for modelling quantum mechanics that is among the most popular tools of the material scientists [Fig. 1(left)]. DFT became practicable with Kohn–Sham’s formulation (KS)39 some 15 years later. KS is the non-interacting Schrödinger equation, which is an iterative procedure minimizing total energy while pairs of electron densities differ. Improving on the computational speed, density-functional tight-biding (DFTB) based on DFT, leverages experimental data40,41. While there is a considerable growing demand to predict material properties increasing in both size and complexity, fundamental run-time complexity of interacting electrons place hard limits on scaling DFT42,43 from polynomial run-time and, hence, DFTB too. Figure 2 illustrates the dramatic bottleneck of run-time as a function of atoms.
(Left) The DFT algorithm. (Right) The cooperative model framework for ({mathrm{TiO}}_{2}) as an example. The first pass (langle 1 rangle) uses DFT/DFTB to produce the minimal viable data (the smallest data size that will produce effective co-modeling). Once the co-model is constucted, we can directly submit the atomic geometry data (langle 2 rangle) for ({mathrm{TiO}}_{2}) to predict Kohn–Sham Total Energy without having to use DFT/DFTB. Throughout the paper we may abbreviate Kohn–Sham total energy to total energy. We use typical ML formalism describing inputs as a vector ({mathbf {x}}), output as label (ell), and classifier as (f:{mathbf {x}}rightarrow ell).
Run-time (seconds) rate of DFTB+ computation as the number of atoms increase. The abscissa show atom numbers and ordinate task names in DFTB+. SCC = diagonalization + density matrix creation, post-SCC processing: eigenvector uniting + energy-density matrix creation + force calculation. The figure is made using ggplot2, ggpubr, dplyr and reshape2 packages in R.
Recently, focused use of ML has shown promise to improving DFT44,45,46 in limited ways, but a general framework leveraging ML is still missing, and no work, that we are aware of, is applied to DFTB. To further demonstrate the potential of this novel approach, we apply co-modelling to anatase TiO(_{2}) NPs used in a wide range of applications47,48,49,50. Figure 1(right) illustrates our approach with various elements subsequently described in later sections. On the first pass (langle 1 rangle) we visit DFT/DFTB once computing Kohn–Sham total energy for a given temperature and ({mathrm{TiO}}_2) nanoparticles (NPs). This is repeated for increasing temperature associated ({mathrm{TiO}}_2) NPs creating minimal viable data—the smallest data that allows us to build our model. The model itself is an linear ensemble of 147 disparate models selected from a maximum cohort of 230 where either poorly performing correlated models are removed. The collective time to select, train, and test the co-model is orders of magnitude faster than relying on traditional DFT/DFTB. Once the co-model is constructed, the atom locations (geometry) can bypass DFT/DFTB (langle 2 rangle) and be directly sent to the model. To show the effectiveness of this novel approach we have build a web service for ({mathrm{TiO}}_2) using this co-model (https://hasan.shinyapps.io/app_1/) that predicts the Kohn–Sham total energy for any temperature in ([0,{mathsf {K}},1000,{mathsf {K}}]). Co-modelling is a novel, data-centric, very fast, and effective hybrid of DFT/DFTB and ML. Our approach is supported by a long-standing observation of the bias of using a single model51. For our work this means using this approach we can always build more accurate co-models than existing single ML models as our results demonstrate as well.
The remainder of the paper is as follows: “Background and related work” section we provide background on KS and an overview of ML and brief description of the members of ({mathscr {M}}). In “Methodology and experimental results” section, we give a detailed description of co-modelling and show its application to the ({mathrm{TiO}}_{2}) NPs. “Summary and conclusion” section is the summary can conclusion. Both code and data are publicly accessible. [https://github.com/hasankurban/Kohn-Sham-Total-Energy-Prediction.git].
The DFTB formalism contains two major contributions [see Eq. (1)], which are matrix elements (the Hamilton and overlap) and the repulsive potentials52. The fundamental idea of the DFTB method is to implement the second order expansion of the Kohn–Sham (KS) DFT where the electronic density ((rho _0)) is calculated from a reference function,
where (rho _0) is the sum of neutral atomic densities. The first term of Eq. (1) represents a Kohn–Sham effective Hamiltonian;
and the second term is related to the energy due to charge fluctuations
where ({hat{S}}_{mu nu } = langle eta _{mu } | eta _{nu }rangle) is the overlap matrix elements. In the DFTB formalism, the matrix elements of atomic orbitals are pre-calculated and stored. Besides, the second order self-consistent charge (SCC) extension is used in the DFTB method due to the dependence of the DFTB Hamiltonian on the atomic charge. The third term is the repulsive potential which is approximated as a sum of two-center repulsions,
with pair potentials based on the respective atom types and the interatomic distance (R_{AB} = vec {R}_A -vec {R}_B.) Typically this problem is solved by a self-consistent approach. Schematic representation of the self-consistency cycle in Kohn–Sham equations is given in Fig. 1(left).
Temperature dependent structures of anatase phase TiO2 NPs have been obtained using molecular dynamics (MD) methods implemented in DFTB+ code53 with the hyb-0–254,55 set of Slater Koster parameters. Thermal equilibrium is controlled by the NVT ensemble during whole simulations. The time step of MD was chosen as 1 fs. The temperature of the NPs was increased by (50,{mathsf {K}}) up to (1000,{mathsf {K}}).
Ellis et al.56 introduces an ML based framework, where the feed-forward neural network is used, to speed up DFT calculations over aluminum at ambient density ((2.699, {mathsf {g}}/{mathsf {cc}})) and temperatures up to the melting point ((933, {mathsf {K}})). The authors show that their model can accurately predict solid and liquid properties of aluminum. Li et al.57 shows the relations between density and energy can be iteratively and accurately learned with very small data and ML models, which have better generalizability, by including KS equations into the training data. KS equations are solved while learning exchange-correlation functional with neural networks which improve generalization. Chandraskeran et al.58 demonstrates ML models can predict the electronic charge density and DOS from atomic configuration information. Instead of learning of a specific representation of the local electronic properties, the authors propose a grid-based learning and predictions scheme which is more memory intensive, but can routinely handle large systems and improve run-times. Brockherde et al.59 perform the first MD simulation with a machine-learned density functional on malonaldehyde constructing more accurate density functionals for realistic molecular systems. Schleder et al.60 gives an overview of ML techniques used in modern computational materials science to design novel materials and explain the present challenges and research problems.
We first give an informal description of co-modelling to prepare for the detailed description for Algo. 1.
To aid discussion we write (tin {mathbb {N}}_0) for temperature Kelvin and ({textsf {NP}} = {((x,y,z),A)|x,y,z in {mathbb {R}} wedge A in {{mathrm{Ti}},{mathrm{O}}}}) for a collection of nanoparticle position and atom type. Using DFT/DFTB we computationally determine total Kohn–Sham energy (E_{ij}) given temperature (t_i) and nanoparticles ({textsf {NP}}_j) (since (i=j) we use only one subscript). The data set has 21 values:
where ((t,{textsf {NP}})) is called the feature set and E the label, building a “best” function f that, given feature values, gives energy using ML parlance. In this work we are interested in investigating how a linear ensemble of disparate ML models performs. We settled on a linear model, since it is among the simplest, best understood, and most widely used. The types of ML models are quite divers e.g., random forests, support vector machines, neural networks, k-nearest neighbor. Detailed discussion of each model is not feasible here, but links to the implementations used in this work are given in the next section. We train every model (M_k) from a set of candidates using default parameters. Models that either individually perform poorly or are correlated are removed. An linear ensemble of models is further refined yielding f
where (alpha , beta in {mathbb {R}}) are coefficients and constant. Using an additional tuning-parameter grid each candidate model is either optimized for its parameters or removed via cross-validation. The simplicity of Eq. (6) belies its power—any value (t in [0,{mathsf {K}},1000,{mathsf {K}}]) and acceptable ({textsf {NP}}) for ({mathrm{TiO}}_{2}) can be determined directly bypassing the traditional DFT/DFTB technique saving time while preserving fidelity discussed in “Methodology and experimental results” section (Fig. 3).
(Inset right) Flowchart of the algorithm used in the self-consistent solution of the Kohn–Sham equations. Cooperative model framework (co-model) that leverages an ensemble of models driven by data. (Top left) are the features and label. We show a sample of the diversity of models: neural nets, support vector machines, random forests. (Top middle) the steps 1–5 using co-modelling for ({mathrm{TiO}}_{2}) NPs. Use DFT/DFTB (1–3) to generate minimal viable data (21 data points), select, train, co-model (4–5).
Presented here is a new approach to predicting Kohn–Sham total energy of nanoparticles by combining a set of single runs of DFT/DFTB with a linear ensemble of disparate ML models we call the cooperative modelling framework (co-modelling)—the results are rapid and accurate. A model’s individual performance criterion considers either (1) Mean Absolute Error (MAE), (2) Root Mean Square Error (RMSE), or (3) R Squared (({mathrm {R}}^2)). Co-modelling consists of five steps:
Compute KS energy for a small number of temperature, nanoparticles pairs. We call this data (Delta) shown in Eq. (5) and Algo. lines 4–8.
Split the data into training data (used to build individual models) and test data (used to assess models quality). We call these data (Delta _{Train}, Delta _{Test}), respectively in Algo. lines 9–10.
Build a set of candidate models from a large corpus of models ({mathscr {M}}). In this work we are utilizing a possible size of over 230 models. A model is ignored if it is either performing poorly or takes longer than an arbitrary, fixed amount of time to complete ((> 1,{mathsf {hr}})) in Algo. lines 10–16.
Of pairs of correlated models remaining, remove the poorer performing in Algo. lines 17–18.
Construct a linear ensemble, possibly refining the set of models further, by tuning parameters or pruning returning the function shown in Eq. (6) and Algo. lines 19–21.
The initial model class (|{mathscr {M}}| = 230) considers all regression models in caret61 which is a popular ML library in R programming language. All models are initially run with default parameter values and become candidates if they complete computation in less than (1, {mathsf {hr}}). To construct the linear ensemble we leverage a popular R package extant in caret called caretEnsemble62. This package relies on caretStack that, from a collection of models, either prunes or optimally tunes parameters. Discussed next are descriptions and some characterizations of the gamut of models used. ({mathscr {M}}) consists of: lm: Linear regression; glm: Generalized Linear Model; lmStepAIC: Generalized Linear Model with Stepwise Feature Selection, gcvEarth: Multivariate Adaptive Regression Splines, ppr: Projection Pursuit Regression, rf: Random Forest, xgbDart,XgbTree, xgbLinear: Extreme Gradient Boosting; monmlp:Monotone Multi-Layer Perceptron Neural Network; svmLinear, svmRadial: Support Vector Machines (SVMs) with Linear and Radial Kernel Functions; knn: K-Nearest Neighbor; rpart: Decision Tree (cart); gaussprLinear: Gaussian Process; icr: Independent Component Regression. We categorize these ML models as (1) ensemble: RF xgbDart,XgbTree, xgbLinear and (2) and non-ensemble models: lm, glm, icr, lmStepAIC, ppr, gaussprLinear, gvcEarth, svmLinear, knn, rpart, monmlp, svmRadial. Unlike non-ensemble models where only a single model is being built, ensemble models63 consist of collections of the same models constructed randomly. More detailed description is provided next.
In RF the ensemble is a collection of DTs constructed with bagging (i.e., independently of each other), whereas XGB builds weak learners (with any type model) in an additive manner—one weak learner at a time. In ensembles, the final decision is given by combining the predictions of all models (“voting”). XGB lets weak learners vote along the way while RF lets all DTs vote at the end after building all trees. Each model is built using a different sample of the training data and, thus, the sampling method is among the factors which determine the final models. Voting strategy is another factor that can change the final prediction, e.g., weighted or unweighted voting where unweighted voting gives an equal weight to the each DT model. RF is quite popular due to its robustness, e.q., remote sensing64, land-cover classification65, network intrusion detection system66, sleep stage identification67.
Error correlation among the trees and the strength of the DTs are estimated over the out of bag data which is the data remaining from bootstrap sampling. The trade-off between the margin which shows how well a single DT separates a correct class from an incorrect class and the correlations between the trees determines how well RF will perform. Breiman68 is the first RF paper and69 gives a review of RF algorithm. Differing approaches to improve RF; weighted voting and dynamic data reduction70, through sampling71, improving data72, with clustering73.
The stochastic gradient algorithm74 improves75, the first gradient boosting algorithms for big data. Extreme Gradient Boosting (XGBoost/XGB)76 is a scalable gradient tree boosting algorithm proven to work well in many areas, e.g., finance77, bioinformatics78, energy79, music80. Unlike RF, the cost function given in Eq. (7) is solved in an additive manner since it is not possible to optimize it in Euclidean space using the traditional optimization methods; it can be solved using second-approximation81.
where (hat{y_i}) represents the prediction for the (i^{th}) data point and
({mathscr {F}} = { f({mathbf {x}}) = g_{q({mathbf {x}})}} (q; {mathbb {R}}^m rightarrow T, gin {mathbb {R}}^T)) represents the trees’ space. K; additive functions’ number, ({mathbf {x}}_{mathbf {i}}); (i^{th}) data point, n;  data size, m;  input variables’ number, t;  iteration number. q is the structure of the trees and (f_k) is an output of an independent tree structure q with a leaf weight. h denotes a differentiable convex loss function and measures the difference between the true model y and the predicted model ({hat{y}}). (omega) is the penalization parameter which is used to tune the complexity of the model and avoid the overfitting problem.
In this section we briefly explain non-ensemble models under two categories: (1) linear: lm82, glm83, icr84, lmStepAIC85, ppr86, gvcEarth87,88, svmLinear89 and (2) non-linear: knn90, rpart91, gaussprLinear92, monmlp93, svmRadial89. Since the models under the same category learn the data according to different strategies, the final models they create are different from each other. For example, although lm can only learn linear relationships in the data, gvcEartch can learn non-linear relationships as well. icr decomposes training data into linear combinations of components which are independent of each other as much as possible. gvcEarth partitions input data into piece-wise linear segments with differing gradients. Linear segments are connected to obtain basis functions which are used to model linear and non-linear behaviors. svmLinear treats the “best” line problem as the maximal margin problem and solves the 2-norm cost functions (least squares error). svmRadial uses radial basis function as kernel function.
lm is Ordinary Least Squares (OLS) under the assumption that residuals distribution is Gaussian. glm searches for the best line by assuming that the distribution of residuals comes from an exponential family klike a Poisson. lmStepAIC, a type of glm, selects the input variables using an automated algorithm. In ppr in an iterative manner the regression surface is modelled as a sum of general linear combinations of smooth functions. This makes ppr more general than stepwise regression procedures. icr decomposes training data into linear combinations of components which are independent of each other as much as possible. gvcEarth partitions input data into piecewise linear segments with differing gradients. Linear segments are connected to obtain basis functions which are used to model linear and non-linear behaviors. gaussprLinear is a probabilistic regression model and a probability distribution over linear functions that fits training data. rpart is a single tree model trained by selecting an optimal attribute (feature) then partitioning data based on the class data for each value in the active domain. This is akin to Quinlan’s C4.5. The optimal attribute is chosen using a variety of metrics such as Gini or information gain where each metric can lead to produce a different DT91,94,95,96. Subsequently the training data points are sorted to the leaf nodes. The algorithm runs until the training data points are classified to an acceptable threshold; otherwise it iterates over new leaf nodes. After building DTs are pruned to prevent overfitting (performing well on the training data, but poorly over test data). Pruning greatly impacts performance95. A neural network is a weighted, directed graph where nodes input/output and edges weights. monmlp is a recurrent network with a monotone constraint which is used to monotonically increase the behavior of model output with respect to covariates. monmlp handles overfitting using bootstrap aggregation with early stopping. In knn, the data itself is used to make a prediction (lazy learning). knn first searches for k-similar objects in the training data with some distance metric and then uses voting. The choice of k can drastically change the model’s prediction.
In this section we first explain how we generate the minimal viable data and then present our novel machine learning framework over the ({mathrm{TiO}}_{2}) NPs and share our findings.
In Fig. 4 illustrates the theoretical supercell data (3-D, 2-D raw data) of ({mathrm{TiO}}_{2}) NPs and some statistical properties of the portion of training data. The process begins by carving ({mathrm{TiO}}_{2}) NPs from a bulk 60 (times) 60 (times) 60 supercell. Next, structures at different temperatures to attain various ({mathrm{TiO}}_{2}) NPs are computed. Figure 4a illustrates three different NPs generated at (300, {mathsf {K}}), (600, {mathsf {K}}), and (900, {mathsf {K}}). Structural data is taken Kurban et al.97 with a detailed description about the initial geometry of the ({mathrm{TiO}}_{2}) NP model. The structural, electronic and optical properties of twenty-one ({mathrm{TiO}}_{2}) NP models were obtained from the density functional tight-binding (DFTB) calculation98,99,100 at different temperatures ([0,{mathsf {K}},1000,{mathsf {K}}]). Each NP consists of 180 O and 88 Ti atoms. In Fig.  4b shows statistical properties of ({mathrm{TiO}}_{2}) NPs used in this study, e.g., distribution, correlation. The input variables are atom, three-dimensional geometric locations of Ti and O atoms (x, y, z) and temperature. The Energy variable is the output variable. We observe that the variables are linearly non-correlated.
The summary of statistical properties of training data. The abscissa and ordinate show the variable values. Red and blue colors represent O and Ti atoms, respectively. Each input feature and bivariate relationships between the input variables are visualized depending on Ti and O atoms.
Let ({mathsf {AC}}, {mathsf {TE}}) stand for an initial atomic configuration and Kohn–Sham total energy. Treating DFTB as a function, total energy is a function solely of atomic configuration shown as computation (8):
Taking temperature into account, we compute the triple (9):
where ({mathsf {AC}}^{{mathsf {i}},t}) is the new atomic configuration using an initial configuration ({mathsf {AC}}^{mathsf {i}}) paired with total energy at temperature t. The co-model is built from 21 different values described in computation (9).
Our initial examination of the co-model itself might lead one to believe that temperature is the principle driver of total energy. This is not the case. The co-model is learning the relationship of the triple not simply temperature as a predictor of total energy. Our results show (including an active web service) that the triple is so well characterized that, given ({mathsf {AC}}^{mathsf {i}}, {mathsf {AC}}^{{mathsf {i}},0}, {mathsf {AC}}^{{mathsf {i}},50},ldots , {mathsf {AC}}^{{mathsf {i}},1000}) and temperature (t in [0,{mathsf {K}},1000,{mathsf {K}}]), it can accurately predict ({mathsf {TE}}^t). The mechanics of the computation are not easy to describe. Underlying most of AI is the recurring conundrum: the model performs well, but how it works is often impossible to explain. Indeed, an active area of AI is explainability101,102,103. Interestingly, work is now taking place that allows AI to completely describe the phenomenon104. At this point, the results show an effective technique to drastically reduce run-time. It remains for future work how, if even possible, we can make some human-interpretable sense of the computation.
Figure 3(top middle) shows the general stops to constructing a co-model that can efficiently, quickly, and effectively predict structural, electronic and optical properties of NPs. The framework starts with generating a minimal viable theoretical data set using DFT/DFTB (steps 1–3). The data is randomly partitioned (step 4) into training and test boosting ((| Delta _{Training}| > |Delta _{Test}|)). Cross-validation is a standard technique to assess the quality of models. Pearson’s correlation is measured among the model cohort allowing elimination of models that are either highly correlated with a better model or perform badly. The final co-model is built and quality measured (step 5). A more detail description of these steps are provided next. In this work, total energy is eV/atom.
Referring to Algo. 1 experimental detail is presented. The construction begins with 21 ({mathrm{TiO}}_{2}) NPs at temperatures ranging over ([0,{mathsf {K}},1000,{mathsf {K}}]) from DFT/DFTB. Training data, (Delta _{Training}), represents 75% of original data and test data, (Delta _{Test}), the remaining 25%. The ratio Ti/O are the same in both (Delta _{Training}) and (Delta _{Test}). In DFTB, the symmetries of the studied NP models were broken under heat treatment; thus, the co-model ensemble and other ML models were trained over non-symmetric NPs105,106. The goodness of the models use three metrics that produce values in [0, 1]: (1) Mean Absolute Error (MAE), (2) Root Mean Square Error (RMSE), (2) R Squared (R(^2)). Metrics (1) and (2) are error metrics; thus lower values are better. MAE and RMSE are distance metrics that, as the model improves, describe the deviation as a magnitude from the actual data (DFT/DFTB) and not the actual accuracy of the model. R(^2), in a regression model, represents the proportion of variance in the input variables (dependent) and that can be explained by the output variable (independent). The higher R(^2), the better model is. tenfold cross-validation is used to assess the quality of the models and the summary is given in Fig. 5.
Observing RMSE, MAE, R(^2) of each model constructed over the training data using Plot Vertical Box-and-Whisker plots. The tenfold cross-validation results of each model using various metrics are shown in this figure. icr and rpart tend to build models with high variance. Considering the experimental results, the best performing ML algorithm on the training data is ppr. rpart has the worst MAE, RMSE and ({mathrm {R}}^2) values.
The best performing traditional ML model is ppr in the training step. The experimental results indicate that the performances of XGBoost algorithms (xgbDart, xgbLinear, xgbTree) and rf are similar to ppr. Although some single ML models such as ppr, xqbDart, xqbLinear, perform well on training and test data, the our framework makes it possible to create more accurate models than any of these single ML models. Any new data may not be well-modelled by a single ML model; the co-model will always perform well. Figure 6 highlights pairwise training model differences: Fig. 6a shows ({mathrm {R}}^2); Fig. 6b shows linear correlations; Fig. 6c shows a Pearson heat map. Figures for MAE and RMSE are provided as the supplementary material for space reasons. After finding the best models among various regression models, correlations among the models, created during the training step, were measured. The results show that lm-lmStepAIC, lm-gaussprLinear, glm-lm, glm-lmStepAIC,glm-gaussprLinear pairs are highly correlated. We only use gaussprLinear, since its performance is better than others over the training data. The co-model is constructed using the remainder of the cohort. Note that there was none that performed poorly over (Delta _{Training}). The performance comparison of traditional ML models and cooperative model over the training data is presented in Fig. 7. The red-dashed line represents the cooperative model and we observe that our cooperative model performs better than the rest of the ML models. For example, over the training data, the RMSE was (1.5 times 10^{-10}) where ppr, the best traditional model, had (1.6 times 10^{-10}) RMSE. Figure 6c demonstrates the relative importance of individual models in the cooperative model. We observe that svmLinear, xgbTree, xgbLinear and xgbDART are the most important models, respectively, in the cooperative model, and svmLinear is the most dominant one. Finally, we present the performance of the cooperative model and other models over the testing data in Table 1. Based on all the metrics the results demonstrate that co-modelling performed superior than all other traditional single models. Our best model is embedded to the web and can be easily tested (https://hasan.shinyapps.io/app_1/).
Determining the models for the cooperative model (a, b) and the details of the cooperative model (c).
The cooperative model versus the classical ML algorithms: comparison of the cooperative model with the classical ML models over the training data set. The red-dashed line represents the cooperative model. We observe that the cooperative model’s performance is even slightly better the most accurate classical ML models.
We have demonstrated that by pairing DFT/DFTB with a novel, data-driven ML framework—and surprisingly a modicum of data of 21 NPs—we can bypass traditional run-time bottlenecks scientists face when using DFT/DFTB alone. Co-modelling builds a linear ensemble of models that accurately predicts the structural, electronic, and optical properties of nanoparticles (NPs). The collective time to build the ML portion is relatively minuscule and can even be done ab initio. Our solution is open-source and only requires a standard hardware—a laptop. We focused on ({mathrm{TiO}}_{2}) NPs to predict the Kohn–Sham (KS) total energy given some temperatures. The diversity of the ensemble is striking: Extreme Gradient Boosting (xgbDart, xgbLinear, xgbTree), Support Vector Machines (svmLinear, svmRadial), Neural Network (monmlp), Random Forest(rf), Linear and Non-linear Regression (gaussprLinear, gcvEarth, ppr, icr), Decision Tree (rpart), K-nearest Neighbor (knn) algorithms. These results are quite promising and the cooperative model has 1 of R(^2) value and almost 0 of Root Mean Squared error (RMSE) and mean absolute error (MAE) over the training and test data. Data continually changes—new, improving, conjectured. The data-driven co-model technique is well-suited in these situations. Future work includes determining what the best and minimal data can be and ensemble, whether instances of this model differ for materials or types of properties predicted, generating data for general use. Additional future work includes elucidating how the co-model is understanding the relationship possibly leading to a different, perhaps simpler, description of atomic configuration, temperature, and total energy. Another intriguing path is to reverse the computation—can we determine the most likely atomic configuration or class of configurations that would give rise to the total energy. Examining other metal nanoparticles must be investigated to ensure this result is not simply peculiar to ({mathrm{TiO}}_{2}). We are also interested in studying extreme temperatures where experimentation is difficult at best. Finally, code and data are publicly accessible and the cooperative model is available on the web.
The raw/processed data required to reproduce these findings can be shared if requested. For questions regarding data, contact Hasan Kurban at [email protected].
Libbrecht, M. W. & Noble, W. S. Machine learning applications in genetics and genomics. Nat. Rev. Genet. 16, 321–332 (2015).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Bhavsar, P., Safro, I., Bouaynaya, N., Polikar, R. & Dera, D. Chapter 12—Machine learning in transportation data analytics. In Data Analytics for Intelligent Transportation Systems (eds Chowdhury, M. et al.) 283–307 (Elsevier, 2017). https://doi.org/10.1016/B978-0-12-809715-1.00012-2.
Chapter  Google Scholar 
Webb, M. E. et al. Machine learning for human learners: Opportunities, issues, tensions and threats. Educ. Technol. Res. Dev. 69, 2109–2130 (2021).
Article  Google Scholar 
Tahaei, N., Yang, J. J., Chorzepa, M. G., Kim, S. S. & Durham, S. A. Machine learning of truck traffic classification groups from weigh-in-motion data. Mach. Learn. Appl. 6, 100178. https://doi.org/10.1016/j.mlwa.2021.100178 (2021).
Article  Google Scholar 
Liew, X. Y., Hameed, N. & Clos, J. An investigation of xgboost-based algorithm for breast cancer classification. Mach. Learn. Appl. 6, 100154. https://doi.org/10.1016/j.mlwa.2021.100154 (2021).
Article  Google Scholar 
Schmidt, J., Marques, M. R. G., Botti, S. & Marques, M. A. L. Recent advances and applications of machine learning in solid-state materials science. npj Comput. Mater. 5, 83 (2019).
ADS  Article  Google Scholar 
Olivares-Amaya, R. et al. Accelerated computational discovery of high-performance materials for organic photovoltaics by means of cheminformatics. Energy Environ. Sci. 4, 4849–4861 (2011).
CAS  Article  Google Scholar 
Tshitoyan, V. et al. Unsupervised word embeddings capture latent knowledge from materials science literature. Nature 571, 95–98 (2019).
ADS  CAS  PubMed  Article  Google Scholar 
Juan, Y., Dai, Y., Yang, Y. & Zhang, J. Accelerating materials discovery using machine learning. J. Mater. Sci. Technol. 79, 178–190. https://doi.org/10.1016/j.jmst.2020.12.010 (2021).
Article  Google Scholar 
Tabor, D. P. et al. Accelerating the discovery of materials for clean energy in the era of smart automation. Nat. Rev. Mater. 3, 5–20. https://doi.org/10.1038/s41578-018-0005-z (2018).
ADS  CAS  Article  Google Scholar 
Butler, K. T., Davies, D. W., Cartwright, H., Isayev, O. & Walsh, A. Machine learning for molecular and materials science. Nature 559, 547–555 (2018).
ADS  CAS  PubMed  Article  Google Scholar 
Wei, H., Zhao, S., Rong, Q. & Bao, H. Predicting the effective thermal conductivities of composite materials and porous media by machine learning methods. Int. J. Heat Mass Transf. 127, 908–916 (2018).
Article  Google Scholar 
Seko, A., Maekawa, T., Tsuda, K. & Tanaka, I. Machine learning with systematic density-functional theory calculations: Application to melting temperatures of single-and binary-component solids. Phys. Rev. B 89, 054303 (2014).
ADS  Article  CAS  Google Scholar 
Zheng, X., Zheng, P. & Zhang, R.-Z. Machine learning material properties from the periodic table using convolutional neural networks. Chem. Sci. 9, 8426–8432 (2018).
CAS  PubMed  PubMed Central  Article  Google Scholar 
Furmanchuk, A., Agrawal, A. & Choudhary, A. Predictive analytics for crystalline materials: Bulk modulus. RSC Adv. 6, 95246–95251 (2016).
ADS  CAS  Article  Google Scholar 
Ward, L. et al. Including crystal structure attributes in machine learning models of formation energies via Voronoi tessellations. Phys. Rev. B 96, 024104 (2017).
ADS  Article  Google Scholar 
Raza, A. et al. A machine learning approach for predicting defluorination of per-and polyfluoroalkyl substances (PFAS) for their efficient treatment and removal. Environ. Sci. Technol. Lett. 6, 624–629 (2019).
CAS  Article  Google Scholar 
Kurban, H. Atom classification with machine learning and correlations among physical properties of ZnO nanoparticle. Chem. Phys. 545, 111143 (2021).
CAS  Article  Google Scholar 
Kurban, H. & Kurban, M. Building machine learning systems for multi-atoms structures: CH3NH3PbI3 perovskite nanoparticles. Comput. Mater. Sci. 195, 110490 (2021).
CAS  Article  Google Scholar 
Wang, Y. & Ma, Y. Perspective: Crystal structure prediction at high pressures. J. Chem. Phys. 140, 040901 (2014).
ADS  PubMed  Article  CAS  Google Scholar 
Xie, T. & Grossman, J. C. Crystal graph convolutional neural networks for an accurate and interpretable prediction of material properties. Phys. Rev. Lett. 120, 145301 (2018).
ADS  CAS  PubMed  Article  Google Scholar 
Ryan, K., Lengyel, J. & Shatruk, M. Crystal structure prediction via deep learning. J. Am. Chem. Soc. 140, 10158–10168 (2018).
PubMed  Article  CAS  Google Scholar 
Li, W., Jacobs, R. & Morgan, D. Predicting the thermodynamic stability of perovskite oxides using machine learning models. Comput. Mater. Sci. 150, 454–463 (2018).
CAS  Article  Google Scholar 
Barnard, A. S. & Opletal, G. Selecting machine learning models for metallic nanoparticles. Nano Futures 4, 035003 (2020).
ADS  Article  Google Scholar 
Pihlajamäki, A. et al. Monte Carlo simulations of Au38(SCH3)24 nanocluster using distance-based machine learning methods. J. Phys. Chem. A 124, 4827–4836 (2020).
PubMed  Article  CAS  Google Scholar 
Mueller, T., Hernandez, A. & Wang, C. Machine learning for interatomic potential models. J. Chem. Phys. 152, 050902 (2020).
ADS  CAS  PubMed  Article  Google Scholar 
Behler, J. Perspective: Machine learning potentials for atomistic simulations. J. Chem. Phys. 145, 170901 (2016).
ADS  PubMed  Article  CAS  Google Scholar 
Behler, J. & Parrinello, M. Generalized neural-network representation of high-dimensional potential-energy surfaces. Phys. Rev. Lett. 98, 146401 (2007).
ADS  PubMed  Article  CAS  Google Scholar 
Bartók, A. P., Payne, M. C., Kondor, R. & Csányi, G. Gaussian approximation potentials: The accuracy of quantum mechanics, without the electrons. Phys. Rev. Lett. 104, 136403 (2010).
ADS  PubMed  Article  CAS  Google Scholar 
Thompson, A. P., Swiler, L. P., Trott, C. R., Foiles, S. M. & Tucker, G. J. Spectral neighbor analysis method for automated generation of quantum-accurate interatomic potentials. J. Comput. Phys. 285, 316–330 (2015).
ADS  MathSciNet  CAS  MATH  Article  Google Scholar 
Wood, M. A. & Thompson, A. P. Extending the accuracy of the snap interatomic potential form. J. Chem. Phys. 148, 241721 (2018).
ADS  PubMed  Article  CAS  Google Scholar 
Shapeev, A. V. Moment tensor potentials: A class of systematically improvable interatomic potentials. Multiscale Model. Simul. 14, 1153–1173 (2016).
MathSciNet  MATH  Article  Google Scholar 
Behler, J. Atom-centered symmetry functions for constructing high-dimensional neural network potentials. J. Chem. Phys. 134, 074106 (2011).
ADS  PubMed  Article  CAS  Google Scholar 
Podryabinkin, E. V. & Shapeev, A. V. Active learning of linearly parametrized interatomic potentials. Comput. Mater. Sci. 140, 171–180 (2017).
CAS  Article  Google Scholar 
Podryabinkin, E. V., Tikhonov, E. V., Shapeev, A. V. & Oganov, A. R. Accelerating crystal structure prediction by machine-learning interatomic potentials with active learning. Phys. Rev. B 99, 064114 (2019).
ADS  CAS  Article  Google Scholar 
Gubaev, K., Podryabinkin, E. V., Hart, G. L. & Shapeev, A. V. Accelerating high-throughput searches for new alloys with active learning of interatomic potentials. Comput. Mater. Sci. 156, 148–156 (2019).
CAS  Article  Google Scholar 
Deringer, V. L. & Csányi, G. Machine learning based interatomic potential for amorphous carbon. Phys. Rev. B 95, 094203 (2017).
ADS  Article  Google Scholar 
Kurban, H. & Kurban, M. Rare-class learning over Mg-doped ZnO nanoparticles. Chem. Phys. 546, 111159 (2021).
CAS  Article  Google Scholar 
Kohn, W. & Sham, L. J. Self-consistent equations including exchange and correlation effects. Phys. Rev. 140, A1133–A1138. https://doi.org/10.1103/PhysRev.140.A1133 (1965).
ADS  MathSciNet  Article  Google Scholar 
Porezag, D., Frauenheim, T., Köhler, T., Seifert, G. & Kaschner, R. Construction of tight-binding-like potentials on the basis of density-functional theory: Application to carbon. Phys. Rev. B 51, 12947–12957 (1995).
ADS  CAS  Article  Google Scholar 
Seifert, G., Porezag, D. & Frauenheim, T. Calculations of molecules, clusters and solids with a simplified LCAO-DFT-LDA scheme. Int. J. Quantum Chem. 58, 185–192 (1996).
CAS  Article  Google Scholar 
Schuch, N. & Verstraete, F. Computational complexity of interacting electrons and fundamental limitations of density functional theory. Nat. Phys. 5, 732–735. https://doi.org/10.1038/nphys1370 (2009).
CAS  Article  Google Scholar 
Lin, C.-C., Motamarri, P. & Gavini, V. Tensor-structured algorithm for reduced-order scaling large-scale Kohn–Sham density functional theory calculations. npj Comput. Mater. 7, 50. https://doi.org/10.1038/s41524-021-00517-5 (2021).
ADS  CAS  Article  Google Scholar 
Jalem, R. et al. Bayesian-driven first-principles calculations for accelerating exploration of fast ion conductors for rechargeable battery application. Sci. Rep. 8, 1–10 (2018).
ADS  CAS  Article  Google Scholar 
Nagai, R., Akashi, R. & Sugino, O. Completing density functional theory by machine learning hidden messages from molecules. npj Comput. Mater. 6, 1–8 (2020).
Article  CAS  Google Scholar 
Allam, O., Cho, B. W., Kim, K. C. & Jang, S. S. Application of DFT-based machine learning for developing molecular electrode materials in Li-ion batteries. RSC Adv. 8, 39414–39420 (2018).
ADS  CAS  PubMed  PubMed Central  Article  Google Scholar 
Gohari, G. et al. Titanium dioxide nanoparticles (TiO2 NPs) promote growth and ameliorate salinity stress effects on essential oil profile and biochemical attributes of Dracocephalum moldavica. Sci. Rep. 10, 912. https://doi.org/10.1038/s41598-020-57794-1 (2020).
ADS  CAS  Article  PubMed  PubMed Central  Google Scholar 
Li, L. et al. Sub-10 nm rutile titanium dioxide nanoparticles for efficient visible-light-driven photocatalytic hydrogen production. Nat. Commun. 6, 5881. https://doi.org/10.1038/ncomms6881 (2015).
ADS  Article  PubMed  Google Scholar 
Simonin, M. et al. Titanium dioxide nanoparticles strongly impact soil microbial function by affecting archaeal nitrifiers. Sci. Rep. 6, 33643. https://doi.org/10.1038/srep33643 (2016).
ADS  CAS  Article  PubMed  PubMed Central  Google Scholar 
Satoh, N., Nakashima, T., Kamikura, K. & Yamamoto, K. Quantum size effect in TiO2 nanoparticles prepared by finely controlled metal assembly on dendrimer templates. Nat. Nanotechnol. 3, 106–111. https://doi.org/10.1038/nnano.2008.2 (2008).
ADS  CAS  Article  PubMed  Google Scholar 
Wolpert, D. & Macready, W. No free lunch theorems for optimization. IEEE Trans. Evolut. Comput. 1, 67–82. https://doi.org/10.1109/4235.585893 (1997).
Article  Google Scholar 
Gaus, M., Goez, A. & Elstner, M. Parametrization and benchmark of DFTB3 for organic molecules. J. Chem. Theory Comput. 9, 338–354. https://doi.org/10.1021/ct300849w (2013).
CAS  Article  PubMed  Google Scholar 
Aradi, B., Hourahine, B. & Frauenheim, T. DFTB+, a sparse matrix-based implementation of the DFTB method. J. Phys. Chem. A 111, 5678–5684 (2007).
CAS  PubMed  Article  Google Scholar 
Luschtinetz, R., Frenzel, J., Milek, T. & Seifert, G. Adsorption of phosphonic acid at the TiO2 anatase (101) and rutile (110) surfaces. J. Phys. Chem. C 113, 5730–5740 (2009).
CAS  Article  Google Scholar 
Gemming, S., Enyashin, A. N., Frenzel, J. & Seifert, G. Adsorption of nucleotides on the rutile (110) surface. Int. J. Mater. Res. 101, 758–764 (2010).
CAS  Article  Google Scholar 
Ellis, J. A. et al. Accelerating finite-temperature Kohn–Sham density functional theory with deep neural networks. Phys. Rev. B 104, 035120 (2021).
ADS  CAS  Article  Google Scholar 
Li, L. et al. Kohn–Sham equations as regularizer: Building prior knowledge into machine-learned physics. Phys. Rev. Lett. 126, 036401 (2021).
ADS  CAS  PubMed  Article  Google Scholar 
Chandrasekaran, A. et al. Solving the electronic structure problem with machine learning. npj Comput. Mater. 5, 1–7 (2019).
Article  Google Scholar 
Brockherde, F. et al. Bypassing the Kohn–Sham equations with machine learning. Nat. Commun. 8, 1–10 (2017).
ADS  CAS  Article  Google Scholar 
Schleder, G. R., Padilha, A. C., Acosta, C. M., Costa, M. & Fazzio, A. From DFT to machine learning: Recent approaches to materials science—A review. J. Phys. Mater. 2, 032001 (2019).
CAS  Article  Google Scholar 
Kuhn, M. Building predictive models in R using the caret package. J. Stat. Softw. 28, 1–26 (2008).
Article  Google Scholar 
Deane-Mayer, Z. A. & Knowles, J. E. caretEnsemble: Ensembles of caret models. R package version 2 (2016).
Krogh, P. S. A. et al. Learning with ensembles: How over-fitting can be useful. In Proceedings of the 1995 Conference, Vol. 8 190 (1996).
Belgiu, M. & Drăguţ, L. Random forest in remote sensing: A review of applications and future directions. ISPRS J. Photogramm. Remote Sensing 114, 24–31 (2016).
ADS  Article  Google Scholar 
Rodriguez-Galiano, V. F., Ghimire, B., Rogan, J., Chica-Olmo, M. & Rigol-Sanchez, J. P. An assessment of the effectiveness of a random forest classifier for land-cover classification. ISPRS J. Photogramm. Remote Sensing 67, 93–104 (2012).
ADS  Article  Google Scholar 
Farnaaz, N. & Jabbar, M. Random forest modeling for network intrusion detection system. Procedia Comput. Sci. 89, 213–217 (2016).
Article  Google Scholar 
Fraiwan, L., Lweesy, K., Khasawneh, N., Wenz, H. & Dickhaus, H. Automated sleep stage identification system based on time-frequency analysis of a single EEG channel and random forest classifier. Comput. Methods Prog. Biomed. 108, 10–19 (2012).
Article  Google Scholar 
Breiman, L. Random forests. Mach. Learn. 45, 5–32 (2001).
MATH  Article  Google Scholar 
Biau, G. & Scornet, E. A random forest guided tour. Test 25, 197–227 (2016).
MathSciNet  MATH  Article  Google Scholar 
Mohsen, H., Kurban, H., Zimmer, K., Jenne, M. & Dalkilic, M. M. Red-rf: Reduced random forest for big data using priority voting and dynamic data reduction. In 2015 IEEE International Congress on Big Data 118–125 (IEEE, 2015).
Elkan, C. Boosting and naive Bayesian learning. In Proceedings of the International Conference on Knowledge Discovery and Data Mining (1997).
Freund, Y. & Schapire, R. E. A decision-theoretic generalization of on-line learning and an application to boosting. J. Comput. Syst. Sci. 55, 119–139 (1997).
MathSciNet  MATH  Article  Google Scholar 
Robnik-Šikonja, M. Improving random forests. In European Conference on Machine Learning 359–370 (Springer, 2004).
Friedman, J. H. Stochastic gradient boosting. Comput. Stat. Data Anal. 38, 367–378 (2002).
MathSciNet  MATH  Article  Google Scholar 
Friedman, J. H. Greedy function approximation: A gradient boosting machine. Ann. Stat. 29, 1189–1232 (2001).
MathSciNet  MATH  Article  Google Scholar 
Chen, T. & Guestrin, C. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining 785–794 (2016).
Carmona, P., Climent, F. & Momparler, A. Predicting failure in the us banking sector: An extreme gradient boosting approach. Int. Rev. Econ. Finance 61, 304–323 (2019).
Article  Google Scholar 
Wang, H., Liu, C. & Deng, L. Enhanced prediction of hot spots at protein–protein interfaces using extreme gradient boosting. Sci. Rep. 8, 1–13 (2018).
ADS  Google Scholar 
Fan, J. et al. Comparison of support vector machine and extreme gradient boosting for predicting daily global solar radiation using temperature and precipitation in humid subtropical climates: A case study in China. Energy Convers. Manag. 164, 102–111 (2018).
Article  Google Scholar 
Murauer, B. & Specht, G. Detecting music genre using extreme gradient boosting. In Companion Proceedings of the the Web Conference 2018 1923–1927 (2018).
Friedman, J. et al. Additive logistic regression: A statistical view of boosting (with discussion and a rejoinder by the authors). Ann. Stat. 28, 337–407 (2000).
MATH  Article  Google Scholar 
Hocking, R. R. Developments in linear regression methodology: 1959–l982. Technometrics 25, 219–230 (1983).
MathSciNet  MATH  Google Scholar 
Nelder, J. A. & Wedderburn, R. W. Generalized linear models. J. R. Stat. Soc. Ser. A Gen. 135, 370–384 (1972).
Article  Google Scholar 
Comon, P. Independent component analysis, a new concept?. Signal Process. 36, 287–314 (1994).
ADS  MATH  Article  Google Scholar 
Efroymson, M. Multiple regression analysis. In Mathematical Methods for Digital Computers (eds Ralston, A. & Wilf, H. S.) 191–203 (Wiley, 1960).
Google Scholar 
Friedman, J. H. & Stuetzle, W. Projection pursuit regression. J. Am. Stat. Assoc. 76, 817–823 (1981).
MathSciNet  Article  Google Scholar 
Friedman, J. H. Multivariate adaptive regression splines. Ann. Stat. 19, 1–67 (1991).
MathSciNet  MATH  Google Scholar 
Zhang, W. & Goh, A. T. Multivariate adaptive regression splines and neural network models for prediction of pile drivability. Geosci. Front. 7, 45–52 (2016).
ADS  Article  Google Scholar 
Cortes, C. & Vapnik, V. Support-vector networks. Mach. Learn. 20, 273–297 (1995).
MATH  Google Scholar 
Dasarathy, B. V. Nearest neighbor (NN) norms: NN pattern classification techniques. IEEE Comput. Soc. Tutorial (1991).
Breiman, L., Friedman, J. H., Olshen, R. A. & Stone, C. J. Classification and Regression Trees Vol. 432, 151–166 (Wadsworth International Group, 1984).
MATH  Google Scholar 
Roweis, S. & Ghahramani, Z. A unifying review of linear Gaussian models. Neural Comput. 11, 305–345 (1999).
CAS  PubMed  Article  Google Scholar 
Hopfield, J. J. Neural networks and physical systems with emergent collective computational abilities. Proc. Natl. Acad. Sci. 79, 2554–2558 (1982).
ADS  MathSciNet  CAS  PubMed  PubMed Central  MATH  Article  Google Scholar 
Quinlan, J. R. Combining instance-based and model-based learning. In Proceedings of the Tenth International Conference on Machine Learning 236–243 (1993).
Salzberg, S. L. C4. 5: Programs for Machine Learning by J. Ross Quinlan (Morgan Kaufmann Publishers, Inc. 1993, 1994).
Google Scholar 
Quinlan, J. R. Induction of decision trees. Mach. Learn. 1, 81–106 (1986).
Google Scholar 
Kurban, H., Dalkilic, M., Temiz, S. & Kurban, M. Tailoring the structural properties and electronic structure of anatase, brookite and rutile phase TiO2 nanoparticles: DFTB calculations. Comput. Mater. Sci. 183, 109843. https://doi.org/10.1016/j.commatsci.2020.109843 (2020).
CAS  Article  Google Scholar 
Aradi, B., Hourahine, B. & Frauenheim, T. DFTB+, a sparse matrix-based implementation of the DFTB method. J. Phys. Chem. A 111, 5678–5684. https://doi.org/10.1021/jp070186p (2007).
CAS  Article  PubMed  Google Scholar 
Luschtinetz, R., Frenzel, J., Milek, T. & Seifert, G. Adsorption of phosphonic acid at the TiO2 anatase (101) and rutile (110) surfaces. J. Phys. Chem. C 113, 5730–5740. https://doi.org/10.1021/jp8110343 (2009).
CAS  Article  Google Scholar 
Gemming, S., Enyashin, A. N., Frenzel, J. & Seifert, G. Adsorption of nucleotides on the rutile (110) surface. Int. J. Mater. Res. 101, 758–764. https://doi.org/10.3139/146.110337 (2010).
CAS  Article  Google Scholar 
Adadi, A. & Berrada, M. Peeking inside the black-box: A survey on explainable artificial intelligence (XAI). IEEE Access 6, 52138–52160. https://doi.org/10.1109/ACCESS.2018.2870052 (2018).
Article  Google Scholar 
Bhatt, U. et al. Explainable machine learning in deployment. In Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency, FAT* ’20 648–657 (Association for Computing Machinery, 2020). https://doi.org/10.1145/3351095.3375624.
Ye, X., Leake, D., Huibregtse, W. & Dalkilic, M. Applying class-to-class siamese networks to explain classifications with supportive and contrastive cases. In Case-Based Reasoning Research and Development: 28th International Conference, ICCBR 2020, Salamanca, Spain, June 8–12, 2020, Proceedings 245–260 (Springer, 2020).
Chen, B. et al. Automated discovery of fundamental variables hidden in experimental data. Nat. Comput. Sci. 2, 433–442 (2022).
Article  Google Scholar 
Minh, H. D. T., Coman, G., Quang, H. N. & Trong, D. N. Influence of heating rate, temperature, pressure on the structure, and phase transition of amorphous Ni material: A molecular dynamics study. Heliyon 6, e05548 (2020).
Article  Google Scholar 
Kurban, H., Dalkilic, M., Temiz, S. & Kurban, M. Tailoring the structural properties and electronic structure of anatase, brookite and rutile phase TiO2 nanoparticles: DFTB calculations. Comput. Mater. Sci. 183, 109843 (2020).
CAS  Article  Google Scholar 
Download references
The numerical calculations reported were partially performed at TUBITAK ULAKBIM, High Performance and Grid Computing Centre (TRUBA resources), Turkey.
These authors contributed equally: Hasan Kurban, Mustafa Kurban and Mehmet M. Dalkilic.
Applied Data Science Department, San José State University, San Jose, CA, 95192, USA
Hasan Kurban
Computer Science Department, Indiana University, Bloomington, IN, 47405, US
Hasan Kurban & Mehmet M. Dalkilic
Department of Electrical and Electronics Engineering, Kırşehir Ahi Evran University, 40100, Kırşehir, Turkey
Mustafa Kurban
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
You can also search for this author in PubMed Google Scholar
H.K.: conceptualization, methodology, writing and reviewing, investigation, software development; M.K.: conceptualization, methodology, writing and reviewing, investigation, data curation; M.D.: conceptualization, methodology, writing and reviewing.
Correspondence to Hasan Kurban.
The authors declare no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Reprints and Permissions
Kurban, H., Kurban, M. & Dalkilic, M.M. Rapidly predicting Kohn–Sham total energy using data-centric AI. Sci Rep 12, 14403 (2022). https://doi.org/10.1038/s41598-022-18366-7
Download citation
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-022-18366-7
Anyone you share the following link with will be able to read this content:
Sorry, a shareable link is not currently available for this article.

Provided by the Springer Nature SharedIt content-sharing initiative
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.
Advertisement
Advanced search
Scientific Reports (Sci Rep) ISSN 2045-2322 (online)
© 2022 Springer Nature Limited
Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

source

Southlake reviews book by grandson of slave in school bearing his name – The Dallas Morning News

Southlake reviews book by grandson of slave in school bearing his name  The Dallas Morning News
source

The Official Jets Podcast: A Conversation with Marty Lyons on the 2022 Jets (8/23) – newyorkjets.com

Host Eric Allen is joined in studio by Jets radio analyst Marty Lyons. The guys talk about the 40th anniversary of the Marty Lyons Foundation, the granting of wishes for children, ages 3-17, diagnosed with a terminal or life-threatening illness (0:20), and Marty’s 36th Annual Celebrity Golf Classic (14:00). They switch gears and discuss the changes the former Jets defensive lineman sees under Head Coach Robert Saleh (21:00), if Marty would have enjoyed playing for Saleh (23:35), and the veteran leadership that Quarterback Joe Flacco brings to the team in Zach Wilson’s absence (27:15). Lastly, they dive deep into the D-line rotation (29:15) and the next step for the young Jets (33:00). Plus, Marty reacts to fellow Sack Exchange alum Joe Klecko being named one of three Senior Finalists for the Pro Football Hall of Fame’s Class of 2023 (37:30).
Host Eric Allen is joined in studio by Jets radio analyst Marty Lyons. The guys talk about the 40th anniversary of the Marty Lyons Foundation, the granting of wishes for children, ages 3-17, diagnosed with a terminal or life-threatening illness (0:20), and Marty’s 36th Annual Celebrity Golf Classic (14:00). They switch gears and discuss the changes the former Jets defensive lineman sees under Head Coach Robert Saleh (21:00), if Marty would have enjoyed playing for Saleh (23:35), and the veteran leadership that Quarterback Joe Flacco brings to the team in Zach Wilson’s absence (27:15). Lastly, they dive deep into the D-line rotation (29:15) and the next step for the young Jets (33:00). Plus, Marty reacts to fellow Sack Exchange alum Joe Klecko being named one of three Senior Finalists for the Pro Football Hall of Fame’s Class of 2023 (37:30).
Host Eric Allen is joined in studio by former Jets defensive lineman Joe Klecko. Klecko reacts to being named one of three Senior Finalists for the Pro Football Hall of Fame’s Class of 2023 (0:20) and explains how he still needs 80% of the vote in January to cement his induction (5:00). They also discuss the Klecko support from Hall of Famers (6:30) and dive deep on what made Klecko such a versatile player (9:15). They recall Klecko’s record-breaking season in 1981 (12:45) and chat about how his amateur boxing career helped him in the NFL (17:01). Lastly, the guys talk about Klecko’s mentality heading into his NFL first training camp as a sixth-round selection out of Temple University (20:35), the story behind Jim Jones from Poland University (21:33), and the meaning behind The Sack Exchange (24:00).
Host Eric Allen is joined in studio by Jets TE Coach Ron Middleton. They talk about his 2-0 record as a head coach after leading the Jets to a win over the Jaguars and guiding the National team to victory at the Senior Bowl (1:30). They dive into the motivation behind his viral pregame speeches (2:00), and how he got into coaching after his NFL playing career (4:15). Coach Middleton reflects on former David Cutcliffe’s influence on his coaching career (6:40), his experience coaching with with Nick Saban at Alabama in 2007 (9:15), and what was his first impressions of Robert Saleh as an assistant coach when they coached together in Jacksonville (14:00). Lastly, they break down the tight ends room, starting with veterans CJ Uzomah and Tyler Conklin (19:05), rookie Jeremy Ruckert (20:58), and Kenny Yeboah/Lawrence Cager/Trevon Wesco (22:20).
Host Eric Allen is joined in studio by NFL Network reporter Judy Battista. Judy gives her training camp takeaways after talking to Coach Saleh post-practice (1:30) plus how does the team take the next step this season (3:05). They discuss the second-year quarterbacks around the league (7:30), why Judy is buying into the Jets’ culture change (10:00) and how important it is for the Green & White to be successful inside their division (13:05). After chatting about Sauce Gardner (17:04), they review the improvement of the defense (19:00), what Mekhi Becton has to do to in Year 3 (25:30), and top NFL training camp storylines (29:00).
Host Eric Allen is joined in studio by Jets Wide Receivers Coach Miles Austin. They talk about the homecoming for Austin coming back to New Jersey (0:30), what he tells his players about his experience in training camp (2:15), and what is unique for receivers in this offensive system (4:15). Coach Austin goes on to explain what “strain” means to the Jets (6:33), how the team is growing up together this summer (8:18), and what is the next step for Elijah Moore (9:30). Lastly, they talk about what it’s like to coaching first round pick Garrett Wilson (11:05), the benefits that this is a skills group that can grow up together (13:30), and what is the difference in Zach Wilson in Year No. 2 (16:05).
Host Eric Allen is joined in studio by SiriusXM radio host and former Jets offensive lineman Willie Colon. They start by catching up on all that Colon is working on both in sports media and reality television (0:30). The guys discuss what is on the horizon for the Jets in Year 2 under Robert Saleh (3:05), Colon’s first season wearing the Green and White (5:20), and what the Jets have been building currently on the offensive line (7:05). Specifically, Colon breaks down what he is seeing from Mekhi Becton (9:05), offensive line play as a whole around the NFL (12:55), and Zach Wilson next steps in Year No. 2 (17:36).
Host Eric Allen is joined in studio by Jets Offensive Coordinator Mike LaFleur. They talk about his close relationship with older brother, Green Bay Packers head coach Matt LaFleur (1:50), the nuances of a run-first offense (4:20), and how Zach Wilson’s athleticism helps him as a play-caller (6:15). The guys discuss the addition of former 49ers guard Laken Tomlinson (8:05), the competition at left tackle between Mekhi Becton and George Fant (10:20), and how the offseason additions can help the development of Wilson (11:45). They dive into the revamped tight end room (13:00), how Coach LaFleur characterizes his play-calling (17:30), and what to expect from Wilson in training camp No. 2 (19:00). Lastly, Coach LaFleur talks about his approach to training camp practices (22:15), his relationship with new Dolphins Head Coach Mike McDaniel (24:10), and his favorite camp memory (26:55).
Host Eric Allen is joined in studio by Jets Defensive Coordinator Jeff Ulbrich. They go in-depth about the type of players the Jets added this offseason (1:30), the importance of winning on early downs (4:15), and the state of the run defense (6:10). The guys discuss the return of edge rusher Carl Lawson (12:30), the rotation favored along the defensive line (15:05), the transition Ahmad ‘Sauce’ Gardner will face from college to the pros (18:20), and the depth at the cornerback position (21:55). They dive deep into the safety position (26:00), how excited Ulbrich is for the upcoming season (33:20), and how the entire defensive staff is united by mindset (36:20). Plus they take a quick peek at the 2022-23 schedule (38:00).
Host Eric Allen is joined in studio by Jets Special Teams Coordinator Brant Boyer. They review the offseason movement that will impact the special teams unit (2:45), All-Pro returner Braxton Berrios’ reaction to his strength rating on EA Sports Madden NFL 23 (3:30), and the upcoming kicker battle between Greg Zuerlein and Eddy Piñeiro (6:15). The guys discuss LB Del’Shawn Phillips’ special teams strides (10:50), the big spring that WR Jeff Smith had (12:10), and the progression of punter P Braden Mann (15:10). Coach Boyer also dishes on the consistent nature of long snapper Thomas Hennessy (16:20), what makes Robert Saleh a great coach to play for (19:50), and how he is already diving into AFC North tape (22:00).
Host Eric Allen is joined in studio by Jets defensive line coach Aaron Whitecotton. They review what was different in the second year of spring workouts (0:20), what changes for young players in moving from college to NFL (3:00), and the additions the team has made on the defensive line this offseason (4:05). Coach Whitecotton looks back at draft night and the picks of Jermaine Johnson and Michael Clemons (6:00). They also discuss what makes Carl Lawson so unique as a pass rusher (10:45), the defensive line mentality and rotations (13:50), and the versatility of John Franklin-Myers (16:05). Lastly, they delve into Quinnen Williams’s next step in his career (17:20), Coach Whitecotton’s favorite Lou Holtz story from his playing days at South Carolina, and his journey from offensive lineman to coaching the defensive side of the ball (21:55).
Host Eric Allen is joined in studio by Jets safety Jordan Whitehead. The veteran DB talks about his introduction to the Jets fans at the MetLife Stadium Draft Party (1:10), why this could be the best secondary Whitehead has played on in his career (1:42) and the talented cornerback pair of DJ Reed and Sauce Gardiner (4:00). The guys also discuss his transition to the Jets defense (8:43), the term box safety (10:07) and how close he has to be with C.J. Mosley to call/align the defense (14:15). Lastly, they chat about how a safety balances going for the big hit or the takeaway (17:51), Jordan’s family connection with former Jets CB Darrelle Revis (20:00) and what it means to be from Pittsburgh, PA (21:33).
Host Eric Allen is joined in studio by former Jets WR Jerricho Cotchery. They start by talking about J-Co being drafted in 2004 (1:15), his rookie season with the Jets (2:30) and his breakout campaign in 2006 (4:45). The guys also discuss what Chad Pennington meant to Cotchery’s career (7:35), what it was like playing for Rex Ryan (9:15) and the playoff runs in 2009 and ’10 (12:50).J-Co reflects on his 2010 play vs Browns when he injured his groin but still made a huge catch in OT (14:30), how tough it was to say goodbye to the Jets organization in 2011 (20:35), and his thoughts on the current team (26:00).
Host Eric Allen is joined in studio by Jets WR Corey Davis. They talk about dealing with injuries in Year 1 with the Green and White (0:53) and how early last season he clicked with QB Zach Wilson (2:37). Davis goes on to discuss his rehab to get back on the playing field (4:13), his spring work out in Nashville with Wilson (4:45), and what the game last season vs. the Titans meant to him (6:37). The guys discuss what type of WR group is building at 1 Jets Drive (9:40), specifically about what he sees in rookie WR Garrett Wilson (1030), and what has changed in the tight end room (12:05). Lastly, they talk about the biggest difference between brothers Mike and Matt LaFleur (14:22), what Davis has seen from Sauce Gardner on the practice field (16:00) and his relationship with head coach Robert Saleh (18:00).
Host Eric Allen is joined in studio by Jets QB coach Rob Calabrese. They talk Year 2 in the system (0:26) and review Calabrese’s offensive coordinator stint at the Senior Bowl (1:00). The guys also discuss the impact that Greg Knapp had last season and his lasting impact (4:45), Zach Wilson’s rookie season (11:10), and how different the QB is this spring (14:40). Calabrese comments on his relationship with offensive coordinator Mike LaFleur (15:15), why this is the right spot for him right now (19:03), and what he took away from his Denver coaching experience (21:45). Lastly, they talk about Calabrese’s high school days on Long Island (24:20), playing both QB and WR at UCF (26:15), and his thoughts on the weapons surrounding Wilson (28:45).
Host Eric Allen is joined in studio by Jets linebacker C.J. Mosley. They begin with thoughts on the Jets’ home opener against Mosley’s former team, the Baltimore Ravens, (0:20) and the Weeks 1-4 run against the AFC North (3:00). The guys also discuss how much instincts play a part of the game (7:00), Mosley’s leadership role (10:00), and the growth of fellow LB Quincy Williams (14:20). C.J. talks about the linebackers he watches around the league (20:45), what separates the successful teams (22:40), and what he likes most about what he’s seeing this offseason on the field (23:40). Lastly, they talk about Ahmad “Sauce” Gardner’s play in OTAs (24:50), the moves that Joe Douglas made in free agency (28:05) and the jump that quarterback Zach Wilson can make this season (34:30).
Host Eric Allen is joined by former Jets QB Mark Sanchez. They reminisce about when Mark was here at the facility to start his career (1:58), his private workout with the Jets before the 2009 NFL Draft (5:22) and the four seasons he played for the Green and White (9:10). Mark shares the story behind his engagement proposal to actress Perry Mattfeld (13:50), what he is seeing from Joe Douglas and Robert Saleh during OTAs (20:30) and his thoughts on current Jets QB Zach Wilson (23:47). Finally, Mark gives his opinion on why Saleh is perfect for this young team (27:45) and dishes on his most memorable stretch playing for the Jets (28:30) including the 2010 playoff victory at Foxborough over the Patriots (32:30).
Host Eric Allen is joined is studio by Running Back Michael Carter. They discuss the difference preparing for Year 2 as opposed to a rookie year (0:35), the veterans in the locker room who’ve impacted him (1:32) and the impact Carter’s family has had on his life (3:15). Carter talks about what he did the offseason (7:20) and his thoughts on GM Joe Douglas calling him “Young MC” (9:55). They reflect on the the mental side of the NFL (14:05) and break down pad level and contact balance (17:15). Carter dishes on the addition of second-round pick Breece Hall (23:52), shares a fun story about a chance meeting with Brian Baldinger an airport (25:07) and his first-guy up mentality (26:10).
Host Eric Allen is joined by former Jets OL/ESPN analyst Damien Woody. They discuss the direction of the team (1:00) and Robert Saleh’s changing of the culture (2:00), the Jets’ standing in the AFC East (3:10) and the potential of the offensive line this season (6:50). Damien gives his take on both the next steps for QB Zach Wilson (8:30) and the improved weapons around him at wide receiver (11:55). They switch to defense and review what the Jets have done this offseason (13:50), the synergy between Saleh and GM Joe Douglas (18:18) and follow with a breakdown of the 2022-23 Jets schedule (19:25). Finally, they turn back the clock and look at why Damien chose to come to play in New York (22:05) and why he decided to stay in the area following retirement (33:05).
The 2022 Jets schedule has been released and host Eric Allen is joined by the NFL’s Vice President, Broadcast Planning, Mike North, to break it all down. The guys discuss the league’s schedule unveiling (01:35) and how it is a combination of both art and science (2:15). They review the decision on the Jets’ one prime-time game (7:25) plus North explains the league’s thought process behind making Week 1 a home opener for the Jets (10:45), if rest differential goes into the thinking of creating a schedule (11:40) and how most teams recently have declined byes coming back from London (14:25). The guys discuss the rare start of the Jets’ slate with the Green & White opening up with four games against the AFC North Division (16:05), the thought process on how the league likes to stack division road and home splits (19:15) and how the increasing number of broadcast partners impacts schedule-making (21:15). Lastly, North gives some insight on the flex scheduling that occurs as the NFL season rolls on (25:30) and some advice for someone looking to pursue a career path in the NFL programming industry (28:40).
Host Eric Allen is joined in studio by NFL analyst Brian Baldinger. They start out discussing the additions made by the Jets via the 2022 NFL Draft (0:30), Baldy breaks down how the dynamic has changed for the offense after the drafted players and free agents (2:30). The guys discuss Zach Wilson and how he does not have to take a leap all by himself with all the new weapons on offense (7:20). Then they switch to the other side of the ball and talk about the additions on defense that the Jets have made during the offseason (12:02), and go deeper looking at the depth at the cornerback position (14:00). Lastly, they discuss the impact that Joe Douglas had made as general manager (15:15) and his relationship with Robert Saleh (17:30).
Host Eric Allen talks with Jets assistant general manger Rex Hogan. Hogan reviews all seven selections the Jets made in the 2022 NFL Draft and how the group will help the Green & White. Hogan also details the Jets’ roster building process during the 2022 offseason.
The final bell has rung on the 2022 NFL Draft and Eric Allen, Anthony Becht, Erik Coleman and Ethan Greenberg put a bow a successful Jets’ weekend. They compare the Jets’ class to other league hauls (0:33), discuss how the Jets found a wideout in the draft (3:43) plus the youth movement inside the locker room (4:50). They take a deep dive into Sauce Gardner’s fit in the Jets system (7:55) and which position group has improved the most (9:05). Lastly, the guys deliver their favorite image(s) of the draft (20:55), dish on if the Jets have closed the gap in the AFC East (24:30) and deliver final thoughts on the Jets’ 2022 NFL Draft (29:15).
Day 2 of the 2022 NFL Draft saw the Jets make 2 selections and host Eric Allen is joined in studio with Anthony Becht and Erik Coleman to break it all down. The guys talk about the selection of Iowa State running back Breece Hall with the 36th overall pick (1:00), They discuss taking a running back early in the draft (2:35) and how Hall could fit in the Jets offense (4:40). They dish on the speed emphasis in the early rounds of the draft (11:15) and how the first four picks are good team fits (14:13). We then hear from the newest Jets running back, Hall, who talked with team reporter Ethan Greenberg (18:00).
Greenberg then joins the set to discuss his thoughts on the Jets’ draft haul to date (21:50). The guys analyze the homecoming for new Jets TE/Long Island native Jeremy Ruckert (23:30) plus the Hall addition to the running back room (25:00). They talk their favorite picks (28:50) and look ahead to Day 3 and best players available (37:30).
Day 1 of the 2022 NFL Draft saw the Jets make 3 selections as host Eric Allen is joined in studio with Anthony Becht and Erik Coleman to break it all down. The guys start by talking about the Jets trading back into the first round to select Florida State Edge Jermaine Johnson (1:00) and later discuss how Douglas and Robert Saleh are on the same page (2:30). We then hear from the newest Jets defender , Johnson, who talked with team reporter Caroline Hendershot in Vegas (4:50). The guys follow that up with a complete breakdown of Ahmad “Sauce” Gardner (10:35) and we head back to Vegas to hear from the 10th overall pick Garrett Wilson (11:55). Lastly, they take a look at options for the Jets ahead in Day 2 of the draft (20:45) and Caroline catches up with “Sauce” Gardner (25:17).
It’s time for final words before the the 2022 NFL Draft as host Eric Allen is joined by Leger Douzable. Douzable delivers a best-case scenario for the Jets at top of the draft (2;00), his impression of Florida State edge rusher Jermaine Johnson (4:30) and how he would decide between selecting an OL or Edge at Pick No. 4 (6:45). The guys talk about the top DBs in the class including Sauce Gardner, Kyle Hamilton and Lewis Cine (9:05). Douzable also gives his take on Deebo Samuel (15:45), the Jets needs heading into the draft (27:30) and his upcoming Sirius Show “I Am Athlete Tonight” (34:02).
COPYRIGHT © 2022 NEW YORK JETS

source

New Multidisciplinary Data Science Master's Starts this Fall – UConn Today – today.uconn.edu

Our websites may use cookies to personalize and enhance your experience. By continuing without changing your cookie settings, you agree to this collection. For more information, please see our University Websites Privacy Notice.
|
Meeting the growing demand for data scientists to tackle the most complex problems in society
(Sean Flynn/UConn Photo)
From online streaming to health care systems, from retailers to researchers, the demand for knowledgeable and skilled data scientists has never been greater.
UConn has stepped up to meet the need with the launch of a new multidisciplinary Master’s in Data Science program and an inaugural full-time cohort of 20 students starting in the upcoming Fall 2022 semester.
“The demand has been nothing short of extraordinary,” says Peter Diplock, UConn’s associate vice-provost for the Center for Excellence in Teaching and Learning and co-director of the master’s program.
Based on the Storrs campus, the 11-month and 30-credit program draws from faculty expertise in five of UConn’s schools and colleges: the College of Agriculture, Health and Natural Resources; the School of Business; the School of Engineering; the College of Liberal Arts & Sciences; and the Neag School of Education.
Beyond the 18 credit multidisciplinary core curriculum – which includes a strong focus on ethics as well as design, programming, machine learning, and data analysis – students can choose from 12 different areas of concentration within the program that range from bioinformatics to cybersecurity to social and behavioral analytics.
“Although the core set of knowledge for students who are interested in applying data science to something like marketing are the same as someone who would apply it to health informatics, the way in which it’s applied, the way you get the data, the way you deal with the data, the way you interpret the data once it’s collected, is going to be different for somebody who’s doing marketing than somebody who’s doing health informatics,” explains Kent Holsinger, UConn’s vice provost for graduate education, dean of The Graduate School, and co-director of the master’s program.
The interest has been extraordinary, and the diversity of students interested has been affirming. &#8212 Peter Diplock, Associate Vice Provost
“We wanted to make sure that students had a good, solid foundation in the fundamentals of data science,” Holsinger says, “with the opportunity to specialize in areas that were of particular interest to them and for pursuing careers.”
All students in the new master’s program will complete a team-based capstone project, where they’ll work to solve a real-world problem and develop practical skills through an experiential learning opportunity.
“We’re cultivating opportunities to work with employers in the state of Connecticut, and in the region and beyond, on projects – messy, sticky problems that are characteristic of the kind of projects employers face daily,” Diplock says. “These projects are critically important for students to be able to develop key competencies of practice related to data integrity and quality, surfacing assumptions, iterative model building, and communicating insights and impacts.”
While the initial cohort will partake in an in-person program, the University plans to launch a parallel and fully online program in Fall 2023, designed for people who are working but interesting in enhancing their skills or pivoting in their career.
“The interest has been extraordinary, and the diversity of students interested has also been affirming,” says Diplock. “When we envisioned this program, we thought about a student who has an undergraduate degree in economics, or in education, or in business, or in fine arts, or in history, or physics or computer science. When I say it’s been affirming, it’s because we’re seen exactly that, where the students who are attracted this program come from diverse prior academic backgrounds.”
Planning for the new program began four years ago, when Diplock – who, in his role with the University, works to take program ideas, research them, shape them, and bring them to life – received two proposals from two different department heads about launching new programs in data science. He engaged a diverse faculty team to explore the concept, to connect with industry, and to really learn what knowledge skills and abilities employers needed from data science graduates.
“It was at a time when there was a lot around data analytics, but data science was really just emerging as a sort of distinct field,” he explains, “and as I came to understand the space better, I became more convinced that it was truly a multidisciplinary space. In our program, the extent to which we have embraced and intentionally sought out a multidisciplinary approach is unique.”
As is the program’s focus on ethics – while all students in the program must complete a dedicated two-credit course in data ethics, ethical concepts have been intentionally woven throughout all elements of the curriculum.
“It’s really vital when people are dealing with data, especially these enormous datasets that are now available and widely used, that they be very careful in thinking about how the data were collected, what biases may have been incorporated into collecting the data itself, and then what biases could arise from employing different algorithms,” says Holsinger.
“Models will by their nature be imprecise,” Diplock says, “and we owe it to people to make sure that our students have a deep understanding of what the ethical implications are of the models that they’re building and to be able to take those conversations head-on, as opposed to sidestepping.”
 
For more information about UConn’s new Master’s (MS) in Data Science program, visit masters.datascience.uconn.edu.

Read the article

Read the article

Read the article
University Communications
communications.uconn.edu
[email protected] | (860) 486-3530

source

‹ Previous Next ›