[00:00:05]
>> Hi everyone. Welcome to the, this week's edition of by the idea, as I said in our series, I am thrilled to welcome a professor who remains out from your you see Today, seminar. He's the Steven l. Millett, chair of chemical and biological agent, and professor of chemistry, biochemistry via physics and by engineering at the University of Illinois.

[00:00:25]
I have Champagne. He deceived his b.s. degree in biology from the University of Science and Technology of China, and PH. D. degree from got that Interesting on the guidance of Nobel Laureate, Dr. Francis Arnold. Dr. You, you see he was a project leader at the dog because a chemical company years won many Awards and Kauto toward creep.

[00:00:46]
If he to such articles and 25 office graduate students that was stuck, supposing Academy careers and most recently, he's also the director of the n.s.a. CIA research Institute for Monica synthesis. Though, Let's welcome again, a professor who is often a seminar resolve to get up and just a reminder to all the audience that if you have any questions, pre-schools among the chatter that you and I reached out.

[00:01:10]
Thanks. Ok, thanks for your nice your question. You know, it's a pity that I cannot the visit to your campus that you in person. And actually I have a many friends you know, on your campus. Most of them are from the chemical engineering Department. But of course, I have an opportunity to meet the with them in other conferences.

[00:01:30]
Anyway, so today many thought about a recent turn of work in the interface of that Intel about its machine learning and automation. As many of you know, putting a knowledge or has made a huge impact to the Society. So here is the sense in the Center is the structure of the d.n.a. molecule.

[00:01:51]
The ever since it's a discovery in the early fifties, everyone wants to harness the power of d.n.a., which I think is really the essence of what I'm about. Technology and he has listed important, acknowledges, developed in part acknowledge that you have to become and yet again at the n.s.a. because in technology, p.c.r. reaction, direct operation and a crisper except for the all the inventors of peace, the technologists got a Nobel Prize in the past a few techies and about technology was initially a prior to the medical area and later agriculture and more recently, industrial area.

[00:02:31]
And of course in order to make about technology work, we also have to develop a technology for downstream processing and a scalpel as well. But there's a very actually interesting comparison between the my crack Tronic industry that we're very familiar with. And in the Baltica industry, both industries had major scientific discoveries in the early Fifty's or the Mike Rice, China straight, the major scientific discoveries semiconductor discovery.

[00:03:01]
Energy technology was they invented in the early sixty's, which is the integrated circuits. And then the whole industry took off actually in later, in the early eighty's. Because of the development of the technology such as a matter of a side than a conductor transistors. And the very largest carrier in the technology analysis.

[00:03:27]
So we all have a smartphone. The computing power of the slabs metaphor already exceeded the computing power of the Super computers we had a few decades ago. And similarly for the Patek industry, they, Our industry has the biggest, the scientific discovery in the early Fifty's, which is the discovery of the double strand at the, in a structure and key technology to recover an International was invented a year early seventy's and the whole industry also took off in the later ninety's.

[00:03:59]
However, if you compare to the Tech industry to the cyclotron industry, the growth is still much slower. And in this manner because of the complexity of the biological system. That's why I think in the early Twenty's, a group of researchers, what inspired by the development of the my crack trying to industry and try to apply the engineering principles from my cracked industry to power outage.

[00:04:31]
Which related to the formation of this entire biology field. And here of course, as a sign of hours I want to point out to people associated with the universe of Illinois played a very important role in the development of the my crack training industry. One professor President who invented the transistor and the other person is a Jack rabbit who invented the transistor and Afghanistan, that India with the sort of the what is the Intel about which here I shall Simply defined as the deliberate a desire of improved on novel biological systems using engineering principles.

[00:05:09]
And this engineering presupposes include the Central Asian, moderate vision for Iraq, and that is our system integration. And those, the principles of War widely used. I can that my correction in history and other engineering disciplines. But they just are not widely used or systematically used empowered yet. And particularly, I think is the purpose argued has been an cycle which actually really accelerated product development.

[00:05:36]
So 6 years ago, my colleagues and I established a Center for Intel about edge at the University of Illinois. In this Center, we try to develop a technology path for color, Illinois politics of entry for the best of our manufacturing. So the technology platform includes foundational technologies for Unisom, great as an engineering, you know, engineering in extra And also we try to develop a competition or choose there where we used to help the design of political systems.

[00:06:11]
We try to leverage in this technology problem to address when the challenge is really to help them substantially. And also we tackle different systems including Microsystems fan systems, or many systems. So this is a high powered Fab, is essentially an integrator robotic assistant. It consists of more than 20 instrument patients, as shown in this diagram.

[00:06:43]
And so in the Center of the power for there's a robotic arm that, that can pick and manage this well plate and move it to any instrument on the Palace for repairs. Think of this, the about as an author with automated a laboratory that we can use this System to do many different applications empowered such as a protein generic has an engineering mind about engineering or simply just has of course winning.

[00:07:14]
And those applications are all very diverse. The idea is to break those. The complex was froze for different applications into process modules, Which will be further broken down into unit operations. Now those unit operations are more general applicable. So we can program those unit operations to create a custom designed to work from 4 different applications that we do have an Operating system that controls all the instrument.

[00:07:44]
And also we try to develop A user friendly interface that we can Use, operated assistant director. And of course, This is a robotic assistant Carol, really sorry to work. But you know, you cannot or think by yourself. That's why we partner is a possible to come by artificial intelligence or machine learning to use with this robotic system.

[00:08:21]
So that we can create a system that can self sync, Telfer to our experience the by the South. And this of course is a challenging, but I think it's not impossible as you were not in a minute. So this is the vision that our came up a few years ago for my research group.

[00:08:44]
And for this audience, of course, I don't need to explain what is a our in the mission earning. Essentially We try to Generate a data using the robotic assistant and then try different term machine learning algorithms and to make predictions. And then what implement those predictions and of charge of validated them, create a New data that we are help to refine the model or even create a New machine learning autos.

[00:09:16]
And this, of course, to work with our irritably. Currently, my group is focused are many of 4 areas. The 1st area is really related to technology development that we try to develop the various proteins near him and about engineering come to US, particularly as the tools for the power foundry.

[00:09:37]
And more recently, we also try to develop a New machine learning tools that can be incorporated with the power foundry. And then we try to leverage these the technology path form to a 4th, we met applications. The 1st application is a very, a 2 drug, a discovery and development in which we try to develop a ice report.

[00:10:00]
Since that about you choose to discover thousands of novel natural parts. Some of them can be Discovered as a New antibiotics or you know and are cancer compounds in a 2nd application were interesting in developing microorganisms such as cell factories or production of chemicals and materials. So this is the area that our talk about her today.

[00:10:31]
And lastly, we are also interest in developing New concern about a 2 to 4 human gene therapy, and also fundamental studies are of the committing structure and dynamics. So in the past a few years, we used to about 5 for a few applications already. As a kind of proof of concept for the 1st a study we did is to develop an automated workflow for the things of our talents.

[00:11:02]
Talent is a very important cost of what you know and even tools. And now this most people use in a Christmas Party for the development of the crisper to afford, you know, editing Hell and what the pastor you know added into. And the problem with that kind of things is that it's very tedious because it used many in our past means that what we did is to create this album in the workflow in this was for all we can actually Choose.

[00:11:37]
We create a Library of those parts and based on the target a sequence, we can choose those parts and then assemble them using the robotic assistant. And they're also to do the Transformation of the assemble, the past, meaning in 3 color and then grow the bacteria. And then I say to pass me to the entire workflow can be finished in 24 hours.

[00:12:04]
It's always a fully automated and as a result, we catch a reduced cost or turn to insist character America. And also, we can see he can, you know, Critter, any tennis that is needed for research. In one collaboration, we use our design talents for the treatment of mitochondrial disease.

[00:12:30]
In this study, we created a balance of those talents and then analyze a function and found a few of them show the increase the specificity towards the target gene over to The wild happening. And then we actually handed our, our engineer the talents of our, to our club reader who did a mouse studies and showed that the, indeed our engineer, the cancer worked very well in the mouse model.

[00:13:04]
So this is a nice example to demonstrate the utility of the high power Fab. You know this into proteins nearing. We also try to use a lot of the system for passive engineering, but you know, collaboration with the media. We created a Library of Powers instead of the pass away involved in the synthesis of art and you know, acid, the, what we did it is that we, this is a Library promotors with different strands and also in terms of with different activities.

[00:13:43]
And then we can create a Library of different combinations of those promoters and in times and we try to find one Pathways that we are given a higher Yield. And this of course, is a still kind of a tedious that's why we wondered whether it is possible to all of it or the whole passed away Engineering process with the help of the machine learning to use.

[00:14:20]
So as a demonstration, we figured out a very simple pathway is the like walking path and direct pass way in on a has a 3 genes. But we try to use it in a promoter's with different strengths in front of for teaching. And, and our goal is to find the pass away with the highest i.q. production.

[00:14:42]
So what we did is that we use the robotic a system to construct a small hybrid of those like opium pathway variants. And then we use the machine learning to hear the 1st very simple is that patient optimization. Coupled with the caution process, we can predict what kind of Combinations of the promoters may your other high May produce more like openness.

[00:15:13]
And then we Designed a New Library based on their knowledge and screened the Library again. And then based on the data, we can make another round of prediction. So basically we can irritate iterator, this whole cycle. Many, many times at least I have it. We have this a for a coaster design, build a test and then cycle, which is the in the water pie, the robotic system.

[00:15:45]
And also the machine learning to use. That's why we called or are automated to as I've shown here in the 1st round. We basically kind of probably scare the whole sequence space and then use the machine learning tools to predict make predictions, and then do the experiments. And then we find more again.

[00:16:08]
So we can do it either for 3 months. And in the end, we are needed to evaluate or less than one percent of all the possible combinations to have a tear and optimized the pathway that produced The highest or like at least a higher ally cooking's. Then one you can do just the random a search of the whole Library.

[00:16:36]
The thing is, this is a very nice demonstration of the combination of Intel about Edge automation and the machine learning. And this is the direction that we had to move to was in the future for many other applications as well. And of course, there's a need for the development of a more sophisticated a machine learning to the emission to patent engineering.

[00:16:59]
We also try to automate the whole Center engine power for but in this case, we use our interference to perturb with Jing suppression The expression of all the genes in a cell. And to the heart of our screening cut, identify those imprudent variants. And then we can repeat the recycle bin this case, of course we don't have a machine learning tools yet.

[00:17:25]
This is much more complex than the pastor of 2 edition. But we didn't spend our lot of time to develop this for the autumn is the work for that consists of the growth of the self in advance of operate. The isolation out the generic it in a, from the South and the creator, libraries or gym preservation and then to the transformation for the buyer has proposed screening.

[00:17:54]
Once we finally improve the parents, we were to the corroboration and use that as the template for the next round of improvement. And so this whole workflow could be finished in less than 2 weeks with minimal human intervention. So with this, the workflow, if we can combine machine learning that we really care accelerator the Mad about engineering effort.

[00:18:24]
But even without the machine running, we just did a 2 or 3 rounds. We were able to find Mutant strata that shows greatly improved tolerance to acidic acid. So we sequenced the evolver genomes and found many of them share common with fissions as many as like $41.00. So it's still very hard for US to pinpoint the exact American that mechanism or the improved or you know that more recently we also expanded our effort.

[00:18:59]
We had To pretend to you, you grant Center Grant focusing are the production of the best power fuse and our products are in the Center. I assembled a team of 20 P.R.'s that really chatted for the developer. The cabinet is for the power foundry into this and you test land aspect for securing the learn aspect, which had to develop a more rationally and operators that were happy used for the better one engineering or in them during efforts and in the Center which had to produce a wide variety of our products engine from organic gases that is a basic chemical to alcohol based products.

[00:19:49]
With this a grant we were we able to upgrade our of our Fab. We even bother a few New liquid handling systems. Because through our past experience, we learned that the liquid 100 is a trickster. Men walk off in the robotic system is a little bit with limiting step.

[00:20:13]
But also we Added the smart automation of integrated hardware and software, much better. And I also added a New tester calculus like the d.n.a. fragment analyzer and the Cell imaging microscope and also the offline and the stack coverage as the testing, as you can imagine is also quite a limiting step there.

[00:20:41]
Of course, the janitor or the data we need To in the remaining time, anyone has about our efforts in the area of Micro yourself actress. It was in the robotic a system and the Intel about it to top. My vision is really try to develop an automated the syringe here and powerful in which we were perturbed, the gene expression on a genome scale.

[00:21:08]
Rather than the traditional approach, it selects only a few Jing habits for genetic manipulation. We want to do it in a Hyper proper manner peculiar on the whole genome. And then we were isolated Mutants that show that improve the variance. And we can do that her Ition and also incorporate those entire beneficial Gintaras into the chromosome.

[00:21:33]
And then we can repeat this Cycle in many, many runs for the Chief cost to enable a have to do it in an ultimatum manner. Because otherwise it would be very labor intensive. And also we want to make it a very flexible so you can work of 4 different are obvious batches of our different products.

[00:21:54]
And ideally, we also want to do it in parallel, because in the robotic system, we can grow the cells in excess well paid to in principle, we can handle 96 different the organisers in parallel. And this of course is the director version concept. And as you know, Director Lucian Is a rapidly growing area.

[00:22:18]
So the initial concept actually was demonstrated in $167.00. But it became a few the after scientist of higher this concept for proteins nearing initially as hair in Red and later for as engineering and more recently, or Marijuana engineering. And particularly in recent years, most of the development in the Revolution in Africa star are the development of a to support, you know, means an earring and Amanda Bynes, an earring.

[00:22:53]
And also it had to enable the developers the continuous operation structure. But with the robotic assist, or what do we do essentially is a continuous evolution. So we don't need to really develop a very sophisticated or kind of method to achieve a continuous of Lucian. We simply can use robotic a system coupled with machinery.

[00:23:15]
And we can already achieve continuous evolution of the genomes. And My former advisor actually gotten a Prize because the she Was the 1st I want to demonstrated the resolution concept for printing hearing. And I joined her lab in 1902. So I basically, my whole thesis was the focused are, are the development of direct erosion tools for cronies in Erie and And now I try to extend this direct evolution approach for, you know, engineering, particularly in combination with machine learning and automation.

[00:24:00]
So there is the scene are mentioned in the beginning and I feel, you know, this New field is just like 25 years ago for that erosion for you. And I'm sure there will be a lot of opportunities in the future. So Here I would just like to introduce a few too, so that we developed or directed a genome evolution.

[00:24:29]
The traditional approach is formidable engineering, either for a standard that our Lucian, which is essentially a serial culture transfer. Or it was a Russian design, basically tried to apply system biology tools to identify those rate limiting steps and then will perform genetic manipulations. And this approach of course, works pretty well, but it's a very time consuming and the slow.

[00:24:54]
And in many cases that we also cannot really identify those now into this gene targets. So that's why in the past 10 years, my group, which had a developer in of those the genome scans in the end to couple that with a has risk winning and automation that we can identify many novel watching targets.

[00:25:16]
So because of hundreds, of course, I cannot introduce all these methods and I mentioned 2 of them. The 1st one is based on the crisper technology. So we can, we call the crisper 8 to essentially we use the 3 crisper proteins which are affordable to each other. And we link the one crisper protein to an activity in the man so that we can operate a target gene.

[00:25:43]
And also we link into another crisper to repression the Met. So we can't directly targeting. And in this Crisper based 13th edition place actually we can perform. He was in this system, we can perform all 3, commonly used the genetic manipulations over expression, the regulation and tradition for a targeting.

[00:26:07]
And particularly we can explore the combinations of these the genetic manipulations. Try to find the optimal Combination As a demonstration of what is the system. We choose the beta carotene pathway at the example. Here I show the Major in times involved in this pathway. And the, what we did that we selected dozens of those in targets involved in the scenes, Isabela carotene, and then we performed over expressions, our recognition and direction on these packets.

[00:26:47]
And the choices create a Library of those different combinations and then screen the Library. And in this case, you know, the company has a color, so it's very easy to do the Hazara screening. And we found the best a mutant strand actually involves the knock out of this rock one gene over expression of a human she wants in.

[00:27:13]
And also the regulation of this yard G 90 Are those the gene targets actually work with US in the gesticulate to give the highest with a carrot in production. And similarly, we use the same method to optimize the protein expression. In this case, we also selected a dozen or so of those chimpanzees that involved in protein synthesis and then perform the operation down regulation.

[00:27:44]
And addition on these selected attack is and we were able to find the best a combination that would, you know, the highest salaries activity. And if the targets are or very hard to predict by the Russian design process. And also, we extended this approach to the genome scale, rather than just, you know, the lack the potential targets.

[00:28:11]
We basically want to try all the 6000 and genes in Sacramento, Riza. So in this case, we can use my career technology to synthesize those current d.n.a. that will make the crisper system target the basic energy. So we create a very comprehensive Library and use this Library to identify New targets.

[00:28:32]
And indeed, we were able to find many personal Jessica Jim targets that are, cannot be easily predicted, or even after we get those habits, we still cannot explain why they worked. And in another approach, we also have Little, I do know skill conditions, but also do it's a very The in this case we also use the crisper technology.

[00:29:06]
We basically for each targeting, we designed the car that will bring the crisper to the target place, packaging. And then we're introduced a double strand break to in addition, we introduced and Already nucleotides, the other contents to a March arms that actually have the same identical sequence as the targeting.

[00:29:36]
And when we introduce that article, it will be integrated into across all chromosome to create A to be future issue in attacking a protein, because that is a b.p. tradition where Casa, open reading from a shift to the tide of protein were basically were being activated. But we can also use the my career technology to things as a Library of those articles, actually as many as 24000 and all those articles that work habits are, or the 6 are the genes.

[00:30:11]
And we can transform this Library to East and then do the hard work of screening for you, Rick. In this case, we also rely on the next generation sequencing technology to identify the beneficial of mutants very quickly. Because the field has a we use is a thinker to sell growth.

[00:30:29]
So essentially we just near to the Cell growth a study and a peek at the one grows, the faster that is to one and then sequence them to identify the targets. So initially, we don't know whether this method of our work or not, we didn't approve our concept.

[00:30:45]
The study with picketer a gene that is Involved in the resistance of this kind of ring. So in the presence of this drug can bring, the Isa cell will be cured. But if with this attacking the gene and one which is a transporter, is knocked out, then the cannot bring, cannot enter the cell anymore.

[00:31:08]
So then the cell line up can survive. And so what we did is that we transformed the Library of what you knock out to the used and then just grow the hybrid of cells in the presence of this drug. And we found a few of them indeed can grow.

[00:31:26]
And then we sequence to them, we found all the targets that the knockout habits are Pinker to this target a gene that means that are out of the 600 genes, we just need to whine, spare and to actually identify the exact target for the company. So encouraged by this result, we use of our technology to improve to industry in point of, you know, type ones that acidic, as it turns out, it was for horns and both are linkers of cell growth in an easy to screen.

[00:32:03]
For in both cases we found novel Gene targets. As you can see here, the op can make a cell grow faster in the presence of those inhibitors. And also even we know, you know, these targets, we still don't understand the molecular mechanism, right? Because of these, our involved impose the translation of modifications, we really don't know what made the cells what Tarrant said to these 2 inhibitors.

[00:32:36]
And we can also, you know, repeat the cycle, try to accumulate a beneficial mutations to make the cells even more hardened to the packet inhibitors. As I show here with the 2 runs of the Revolution, and we were able to improve the tolerance very significantly. And as I mentioned, one unique feature of this method is really is the precision.

[00:33:02]
So we can actually Kenya a single nucleotide in the hope Cross up. And the way we do it, it actually is a really depends on how you designed it, or the organ you could have used to do the integration. So we can Insert, or as a few nucleotides to chromosome or just chance, or a single nucleotide.

[00:33:27]
To demonstrated this concept, we pick up the packet, a protein that is involved in the improve the tolerance to 440. And we want to find out what the mutations, What the rescuers are really important for the improve the forefront parents. So we selected this 29, I mean as the region and then change every, I mean I said to 20 different to 19 to the other 19.

[00:33:54]
I mean 0 as it's used in the Chancery method I just mentioned. And then we can analyze the function of the we've had and we've found a few mutations that are responsible or they improve the currents for this to a will be very useful for studying those proteins.

[00:34:19]
There are important for Biggest out development of differentiation because you cannot do a simple back out then the cell when are survive. Or we can use this method to basically introduce site specific, a change into those proteins. Then we can study their function directly. So, microorganisms in a can be engineer by about engineering to use to produce many different chemicals and the materials.

[00:34:54]
However, the bench of products that are what I wanted you can produce is a still very small compared to chemical Callistus country. We have a huge, a chemical industry that produce all kinds of chemicals and the materials. Because if you compare the UK tell US this with a chemical callouses, you know, based on medical list of organic calloused, the chemical canister usually has a much higher productivity.

[00:35:25]
Or it has poor selectively compared to in times or even the microorganisms. Well practiced. And they can have a very hostile activity, but the productivity typically is not a very high. So in the past 10 years, we also really try to combine the energies of both chemical calloused and the power of catalyst.

[00:35:53]
And try to Make those Callista work in the same reaction vessel. And as I just mentioned in the, in the Times has of then are just like the high selectivity. And also they are more amenable for protein cheering by Russian. These are all about a Russian approach, but the problem is, the limitations of that are they are not as stable as the chemical calloused as they don't work at the very high temperature and pressure.

[00:36:28]
And also they have a limited assumption scope and they are not a compatible with the organic solvent. And in many cases, you know the substrate, there are not a soluble in water. So we have to use organic some of our chemical has the carrots, the typical to have a very hire activity because they operate at high temperature pressure and has a broader specifically and was, you know, getting some and of course.

[00:36:55]
But the limitation, as mentioned, is many of the Selectivity to because it's not very high. And also the requirement of a conic organic solvent actually is also in the tissue to there create a lot of environmental hazards. So then we really Wonder, there is a possible to come by these advantages of this 2 type of a House.

[00:37:23]
And particularly we want to achieve the core operative effect because otherwise, you know, you can just run this to ration separately. You don't need to really put all them in the same reaction vessel Because that's a much more challenging, in many cases, that chemical calloused are not compatible with the, in the catalyst of the quite different reaction conditions.

[00:37:51]
But in the past 10 years, actually we developed a number of those tools that can be used to come by in some callouses, with chemical callousness. And of course, is due to timing to our now go through all this examples. This actually we try to come by and has reduced thousands of with in fact and assess the come by a homogeneous callouses, the with Host outweighs the Tarsus audience on task as a leader.

[00:38:20]
That aware of our talk about is the combination of folks houses with INS and tell US so in the 1st example I want to highlight is the combination of photojournalist, catalyze relation ration with an insight to produce that inventor pure company. The concept is the Street here. So we used to the photo catalyst to catalyze that I summarization of this arsons which will produce a mixture of isomers.

[00:38:56]
And then we use the inside in practice to cover the wild isomer selectively to the desired in then top your product and the in the one hour counter. And I swear that we can achieve a very high in until selected. And the many chemical approach that are being developed for all of the instances, such as the No vengeance gel a concession, a way to get out of the Nations.

[00:39:27]
But all this process that we are, that are to the a little to the formation of a mixture of those isomers. And this, I suppose, cannot be the lack to recover that by no great artist has a show here. Either the result is available in the church or cannot work on a subsidy or, or can at the most or you know what I want to achieve and 50 percent and I hear God.

[00:39:57]
All you can achieve are 100 percent Your And then there's no in that you're selected. And we said after listening with artists, because It is a very robust, Calloused, and a very versatile, it were used to the f.m. cofactor to reduce the number one in the helping US to produce and in n.t. or pure product.

[00:40:24]
And this instance have several advantages, such as a work at or deny reaction conditions and also show that very high in and just activity or while for other products. Also, it shows a high tolerance to organic solvent and elevate temperature. And to demonstrate our concept, the key is cost to make the photo canister and the in terms of working together in the same reaction vessel.

[00:40:53]
There are several requirements for the focalized as a militia in action. The ration must happen in the Crystal Ocean because of you know, that in the washing is a solution. And also as a militia in Russia, the faster than the radical reduction reaction. So that the coverage of a could be achieved.

[00:41:16]
And also we want to make sure the Forecaster it's compatible with inside and also the cofactor regeneration system. So we screened many different for the catalyst. And we found that the radiant based the full calloused and also the phone and, and the bright work at the best. It was a, achieve the highest Conversion for the assimilation reaction.

[00:41:46]
And also they water tours are many different subject. And then we also screen the many different in with artists as well, 4 and a substrate to visit, which is a lot of a try Everest alloca magic there. So history that we can try so many in times we can try so many calories can chart that.

[00:42:08]
There's a lot of it that really point out of the need for machine learning tools and our holiday in the later. But here we try that in a many different substrate and unfortunate for all the substrate that they all work about, well they can achieve very high and very high yield.

[00:42:26]
But of course I didn't report as a failed experience for ourselves. Astrid misses it all work and also we have a very, very poor, you know, a set activity for the Chief cause the we want demonstrator, the copper to, you know, between these 2 have a calloused, So as a demonstrator, you know, these 2 reactions for the 2 different substrate, if you run the assimilation reaction and interacting with the action separately, the Yod for this service trade is only 71 percent and are 52 percent.

[00:43:01]
The sum the one percent. While if we put a 2 catalyst in the same reaction vessel, we can achieve 99 A percent, almost 100 percent innocence, activity and 94 percent A year out. And similarly for this substrate the year others are only 11 percent for the 2 Russians operator separately.

[00:43:28]
But if we can combine the 2 colors together because she is 79 percent a year old and 92 percent. So this 2 examples, a clearly demonstrated accord with effect between this the 2 type of calloused. And of course, we ought not just Wonder in our show that our discussion works.

[00:43:51]
We also want to use this tandem callouses system to produce a wider variety of power active come past, including this 2 druggists, the back of a and the fin about to really demonstrate the importance of this tandem. Alice's idea. And in another example, we also Wonder whether it is possible to make US in sand to do both for the calluses and in the America tell US.

[00:44:25]
But in the previous example, where basically did a sequential, not sequential Use the 2 different the catalyst. So here we want to get a possible to really Use a single catalyst to catalyze the 2 types of Alice's. So for the catalyst, I've been widely used for cofactor regenerations for many years.

[00:44:52]
And in the previous example, we show that the 4 keris can also cause as our nation reaction, followed by the in the medical reduction we can achieve carpet in effect. And how has the from Princeton he also showed that You can use the cofactors in the natural occurring of the, with artist to catalyze the photo reaction and followed by that in America reaction to produce some inventor up your camera.

[00:45:26]
But also our interests are a interim interest in intermolecular Rakshas health. So in nature, the Flamen dependence in times like a rat, is well catalyzer to lug around with actions to produce is the inventor of your product. So we Wonder, is a possible we can we propose that in regards to catalyze the some novel reactions figure this one electron Cross coupling reaction to that or to what exists or could be, can test to form a Cross coupled product to our habits or for our company is this After Hello or can your group and also this King And we wanted to repurpose of mind.

[00:46:27]
And in result is that a chair catalyzed the C. C. Banner formation between these 2 substrate to form is the intention of your product. So this reaction has never been reported by any interest. So this is New to nature in America reaction. And also this reaction is a very hard to do chemical because into the, it's a for chemical because it's very hard to achieve the intention or selectivity as I just mentioned before.

[00:47:00]
So this actually basically is a totally novel reaction. And a bizarre understanding of the mechanism of the India and I don't go through that House and we figure out How to rig a cycle for this. In that we think that it is possible to repurpose this in with our pace to catalyze this New to nature reaction.

[00:47:25]
And of course, you know, we scream that many in times. Right? Because we don't know which in our work, and also we screen them any substrate. And that once again is a point out, the need or mission and in tools. And we screamed many as a show here.

[00:47:42]
We actually identify that many of US in real life is short or high, you know, just an activity or some actually Have a higher, you know, there's was some low, right? And sometimes I was even have the opposite. The selectivity and The, you know, there's already the law, but here on it we show a small set of INS and we screen and also a small set of stuff, a Street that we tested here that we did.

[00:48:12]
It has to know why the brightest on Street such as the after Hello Carol ketones after Hello M.R.I.'s and asked US not help things. And we showed that they can achieve a pretty high And also decent yacht. But once again we are, we reported a success for is out there a lot of fail the results.

[00:48:34]
And we did a lot of work in a to identify this substrate or scope. That is actually really data to my, you know, motivation to use a high tooth for chemical Callistus. And I was very fortunate to attend this and as a, our research into Grant a few months ago.

[00:48:57]
So in this New Institute we call the monitor make a lab Institute. Basically we try to develop a New a R 2 and also New chemistry to radically accelerate and democratize since and also functional material discovery as well. So in this Institute, we have for research a Trust in the 1st Trust which has a developer they are in, they will since is planning to use, as are shown in the tandem callouses example, we don't know what are coming in that we should use.

[00:49:30]
We don't know what kind of rational use these are possible to use a high to say to figure out what as with the insulin. What are the Callista, What are the cause? That's the combination of the insides and the chemical calloused that have really produced the functional markers of what with that is either a yacht and Selectively.

[00:49:54]
And also we try to develop a are enabled to support catalyst environment. Here of course is not just that you need an catalyst development, but also is a chemical catalyst. A lot of chemical has to have been used and the need to be optimized managed by another. There are That is the reaction, conditions, McAlister, themselves.

[00:50:15]
And we want to use those Ai to say and also to the robotic assistant I just mentioned is there are about that. And in the Sion in the Center, we also developed the order with a synthesizer for chemical reaction as well. Not just for the power callousness of our logical systems.

[00:50:35]
And I was which as a developer, Atlanta make a database that are watching copper, the data that are generated from the all, with the things eyes are. And I have our fair With the data in the literature to our mind. The lyrics are very comprehensive to get all they need for those chemical reactions.

[00:50:54]
And we want to use these tools are 2 things as a few important target happens and also to discover Use for a monitor use for mature US, particularly For the both hacked materials. And of course, in the sense that we also have very expensive comprehensive Education, Workforce Development, our chip program, we have an innovative programs try to involve many of US.

[00:51:29]
So this is of course we just launched on a 2 month ago. And With just the work Proposed project thought in summary, what I thought today is that we developed And it was our power for, for the next generation to know about applications. And I showed that we can use this system for proteins hearing and a path to engineering or human genome engineering.

[00:51:59]
And I clearly see that's a lot of opportunities for the development of New machine learning tools for not just the powerful but also for our system engineering at a different or a levels. And the final, our data Center, the funding source, mostly from Gilead for most of my projects and also an s.f. In addition, or I stunk it up and I each and also various companies or the different projects in my lap.

[00:52:32]
Thank you for attention of we have better questions. I get so many can't hear one clapping or so he hears one question. So is there a risk that using animal models might cause missing an important substrate? What A repeat again. So the question is, is there a risk that using machine learning models might not think something like missing, like an important substrate which you may want to explore or that separates a possible?

[00:53:07]
Yeah, Right now with many use machine learning for optimization, we haven't actually really developed the machine it was that can help US design the system just like the to some of the substrate or The catalyst. But as I see the need does that stick there? And I point out right at the Door, we publish those papers in a high profile journalist, but we did a lot of work to achieve that.

[00:53:37]
If we have the missionaries was, I think we can really expand that occasion of those concepts. Are there so many focused out there are so many Indians out there. And also the fact is that we really don't know what kind of a reactions we can achieve with this kind of an I Pad of lives.

[00:53:57]
So we have time for a couple of questions. So I had one, I mean, so in your molecule, we collaborate, speaking as a machine learning the person like what kind of data sets do you have access to, which if I want to think in terms of what kind of problems these create as a machine learning problem, right, so what would you suggest like, so do we have like majors or do we have experiments to make them structured data?

[00:54:22]
So how would you put Your benefits? I guess. Yeah. For a sense of panic like in the 1st, the ones who are use all kind of data like that the, the graph for images And of course the literature. And that's why we have experts like a ****. He's a really expert in image analysis.

[00:54:44]
And I also have our experts are in that for language processing and the test mining. And so we basically try to my all the data out there in the chemical iterators. And then in addition, you know, we will use the automated system. We have those, I think our competitive advantages compared to other our competitors.

[00:55:03]
Maybe that has, we can generate a lot of data for our target system. And then we can incorporate those data with the data that are not, you know, literature either in the literature or the patent in. And also we are now having gauges, industry partners, as well as the also, especially from a few companies.

[00:55:22]
They also have a lot of data the for the different have are for Telus. So are there are, there are connections to the next accelerating drug discovery, which a lot of people are also nowadays getting interested. And Yeah, it's definitely, that's why we actually were pushed by many firms to companies already there.

[00:55:41]
Many of them is a more interesting in the Chemical things is part that's for the drug or Habitat. If a kitchen that's a different to US, but once you have the Compound identifier or design, then you want to synthesize it. And that I think this is the focus for this Institute.

[00:56:01]
Really try to come up for the most economical or faster, faster Route for the single system of the target. A company interesting are greater so are there is a question we just have To be all Ok so and I think there also are creepy. I'm exactly on the star Speaker again and I mean again, thank you for taking the time.

[00:56:28]
Very interesting stuff and looking forward to the results from the state. Ok, Great. Thank you very much. Bye bye.