Transcript

Biocatalysis: Sustainable innovations in catalysis for a greener future

From Cell Press · Episode page · Watch on YouTube

· 1 h 38 min · 12,465 words · 98 timestamped sections

This is the full, timestamped transcript of "Biocatalysis: Sustainable innovations in catalysis for a greener future", an episode of Cell Press published November 25, 2025 (1 h 38 min, about 12,465 words). It mentions 15 people, works and ideas, including Shelley Mir, Heming Chao, Jeff Van Drew, Shiao Shiao Chao, National Science Foundation.

Who speaks in this episode?

What is discussed, and when

  1. – 10:00: Shelley Mir, Jeff Van Drew, Shiao Shiao Chao
  2. – 20:00: JAX
  3. – 30:00: Jeff Van Drew, Chinese Academy of Sciences
  4. – 1:00:00: Heming Chao, National Science Foundation, University of Illinois at Urbana Champagne
  5. – 1:20:00: Lycopene, AI Institute, Illinois Biological Foundry, Global Bio Alliance
  6. – 1:30:00: AI Institute

Notable quotes from "Biocatalysis: Sustainable innovations in catalysis for a greener future"

  • Catalysis for a greener future. I am Shiao Chao. I'm a scientific editor at campalysis cell press and I'm very delighted to serve as your moderator today together with my…
    about Shiao Shiao Chao · at
  • With that, it is my great pleasure to introduce our first speaker of today, Professor Shelley Matier from the UN Misery University of Science and Technology, where she serves…
    about Shelley Mir · at
  • Director of Tenjin engineering research center of bioc catalytic technology affiliated with the tenjin institute of industrial biotechnology of the Chinese academy of sciences…
    about Chinese Academy of Sciences · at
  • Finally I will thank my collaborator and my team and also thank the funding from the national science foundation and also the fun founding from the industry partners.
    about National Science Foundation · at
  • Get to them at the end of the end of the talks. and now I would like to move on to Professor Himing Chao. from Ding's last slides, we can already see that automated production…
    about Heming Chao · at
  • And so about 10 years ago we built the integrate robotic system called the Illinois biological foundry for domestic bio manufacturing or IBO fab.
    about Illinois Biological Foundry · at
  • A few years ago and this is a proof of concept study. We try to use multiple enzymes to synthesize an important chemical called lycopine and we express those enzymes in the…
    about Lycopene · at
  • As well. So and this proof of concept study helped me to obtain a grant from NSF in 2020 to establish one of the first seven AI institute in the United States called a molecule…
    about AI Institute · at

Who and what is mentioned in this episode?

People

  • Shelley Mir — Person · Main speaker (panelist) · 3 mentions · first at
    Delivers the first technical talk on bioelectrocatalysis, covering enzymatic fuel cells and enzymatic electrosynthesis, and frames key scale-up challenges.
  • Heming Chao — Person · Main speaker (panelist) · 2 mentions · first at
    Discusses expanding directed evolution via AI-driven automation, biofoundries, and photoenzyme catalysis; references work connected to Francis Arnold’s Nobel-recognized directed evolution.
  • Jeff Van Drew — Person · Main speaker (panelist) · 3 mentions · first at
    Presents industrially oriented biocatalysis work: engineering ThDP-dependent enzymes for aldehyde coupling and microbial biotransformation of phytosterols to steroid intermediates.
    Pioneering work continues to expand what analysis can achieve. We have professor Shelley Mir, Professor Duming Drew and Professor Huming Jaw. Their insights will help illuminate both the current landscape and exciting directions ahead.
  • Shiao Shiao Chao — Person · Host / Moderator · 1 mention · first at
    Introduces the webinar, moderates the session, and introduces each speaker and the Q&A segment.

Companies & orgs

  • National Science Foundation — Company · Funding organization · 1 mention · first at
    Provides funding for research centers and projects discussed in the webinar
  • NSF Center for Synthetic Organic Electrochemistry — Company · Research center · 4 mentions
    Focuses on making organic electrochemistry mainstream
  • AI Institute — Company · Research organization · 2 mentions · first at
    Focuses on developing AI tools for molecular synthesis and discovery
  • Illinois Biological Foundry — Company · Research facility · 1 mention · first at
    Automated lab for rapid prototyping of biological systems
  • Chinese Academy of Sciences — Company · Research organization · 1 mention · first at
    Affiliation of one of the speakers, involved in biocatalytic technology
  • University of Illinois at Urbana Champagne — Company · Academic institution · 1 mention · first at
    Affiliation of one of the speakers, involved in synthetic biology and automation
    L Steven L Miller chair of chemical and biomolecular engineering and the professor of chemistry, biochemistry and bioengineering. when he has led an illustrous career in…
  • Global Bio Alliance — Company · Collaborative network · 1 mention · first at
    Consists of members around the world focusing on biofoundries
    And this effort led to the formation of global bio alliance in 2019 which currently consists of more than 40 members around the world and also many actually in industry and also led to the.
  • JAX — Company · Scientific journal · 1 mention · first at
    Journal where research is published
    So at the time Phil Baron and I had both just started as associate editors for Jax. We met at a meeting and he started quizzing me about sort of electrochemistry with his focus on kind of electrifying.
  • Cava Trust — Company · Funding organization · 1 mention · first at
    Provided initial funding for the development of biofoundry technologies
    Actually achieve autonomous experimentation and we studied this effort in 2013 with the grants from Cava Trust and then you know obtained the grants from various funding agencies including DAPA, NIH, NSF, DOE so basically.

Products & tech

  • Lycopene — Product · Chemical compound · 2 mentions · first at
    Example of a chemical produced using AI-driven pathway optimization

Concepts & topics

  • Suzuki Coupling Reaction — Concept · Chemical reaction · 1 mention · first at
    Example of a reaction optimized using AI tools
    Engineering of catalyst. And here we used the four kind of case studies. one is related to Suzuki coupling reaction. try to optimize the yield of the reaction and other two are…

Full transcript with timestamps

Hello everyone. Welcome to today's cell press webinar biocatalysis sustainable innovations in catalysis for a greener future. I am Shiao Shiao Chao. I'm a scientific editor at campalysis cell press and I'm very delighted to serve as your moderator today together with my colleagues from cam chemcularity and cell reports physical science as balysis is rapidly sh reshaping the frontiers of modern chemical synthesis by offering sustainable highly selective and energy efficient pathways. Now further empowered by AI accelerated in them discovery and optimization, the field is opening for uh for transformative opportunities across chemistry and biotechnology.

It is our great honor today to be joined by an exceptional panel of leaders whose pioneering work continues to expand what analysis can achieve. We have professor Shelley Mir, Professor Duming Drew and Professor Huming Jaw. Their insights will help illuminate both the current landscape and exciting directions ahead. Uh we hope this session will inspire new ideas and conversations. Please feel free to submit your questions at any time. We will address them during the live Q&A. With that, it is my great pleasure to introduce our first speaker of today, Professor Shelley Matier from the UN Misery University of Science and Technology, where she serves as director of the Kmer Institute uh center for resource sustainability and the director of the NSF center for synthetic organic electro Chemistry. Professor Mntir has published nearly 500 research articles and has received numerous awards recognizing her seminar contributions.

Today she will share some of the most exciting recent discoveries from her laboratory. So Shelly the floor is yours.

Thank you. Yeah. So I will stop sharing here work on opening my slides. We'll give it a second for them to come up hopefully. So um thanks for the the very kind in invitation um to to present today. Um I'm an electroanalytical chemist uh who has used bio biocatalysis throughout my career. Um, as an electroanalytical chemist, you might think that today I'm going to talk about bioensors. Um, because I think bioensors are kind of the the classic example of where an analytical chemist uh combines electricity uh or electrochemistry and biocalysis. But uh today I'm going to tell you a couple of stories um in other areas of electroynthesis.

um specifically maybe if I can get my slides to work specifically uh two areas uh one focused on using biocatalyst uh to generate electricity uh so basically energy conversion or energy storage devices uh if the bio catalyst is an enzyme we call that an enzyatic um fuel cell if the biocatalyst is microbe. We call it a microbial fuel cell, but basically using um either a protein catalyst or a whole living organism um catalyst to actually um generate electricity from biological fuels or or from a a organic molecule.

Uh the second side of it uh is using electricity um to actually generate molecules of interest. Um and so these would be sort of electrochemical uh synthesis systems or what we call electroynthesis. Uh so with an enzyme we call that enzyatic electroynthesis. With a microbe we call that um microbial electroynthesis. And I'm going to talk about both of these um today because they've been areas of interest uh in my group um for quite some time but with a focus uh a biocatalysis focus um today um on enzymes uh and not uh necessarily on the microbial side of things.

Although we work a great deal um in the the microbial side. Um having a little challenge getting it to go to the next slide. So on the energy side of things, uh if you think about sort of traditional um fuel cells, uh the most common fuel cell is the hydrogen oxygen fuel cell. um and it basically oxidizes the fuel hydrogen uh separates it into electrons and protons. The electrons go through the wire and to work the protons um cross the membrane react with oxygen um from the air um to form water. And that's the reason they're considered um green um and uh clean um because they're only producing water. Um and this system works really good if you're talking about simple molecules like hydrogen, methane, methanol. um but we don't really have the ability to develop traditional electrocatalytic materials that break carbonarbon bonds effectively. So if you want to work with a fuel like ethanol that has a

carbonarbon bond uh you're going to have challenges with traditional um fuel cell um chemistry. Um so what my group does uh is basically use biocatalysis uh and the ability of sort of metabolic pathways um to do um deep oxidation of fuels and sort of interfacing um those uh enzyme cascades with electrodes um for energy conversion devices. An example I'm going to show you um today is the KB cycle or the citric acid cycle. Um so you can think about uh this uh as a uh as a metabolic pathway in the uh mitochondria of a living cell um that has a whole cascade of oxidtored reductase enzymes. So those are the bright red enzymes uh and uh corresponding non-dehydrogenase um enzymes that will completely oxidize uh um pyuvate to produce um carbon dioxide through both oxidation um and breaking chemical chemical bonds. At the point in time we got interested in biouel cells.

There were pyuvate um bofuel cells out there that had used pyuvate dehydrogenase. But we were interested in sort of seeing whether or not we could use these um catalytic cascades or deep oxidation um to be able to um produce more electrochemical energy and to produce that electrochemical energy faster. Um so uh if we look at what a a fuel cell typically um looks like uh you typically um will have uh a oxygen reduction side um that's going to um basically um take oxygen um from the air and then you have an anode side where we can actually titrate uh the proteins of the citric acid cycle um to actually look at their electrochemical performance. And if we do that, uh, next slide, maybe I can get it to work. Uh, if we do that, you can see, um, that what happens is as we add additional, um, dehydrogenase enzymes and their corresponding non-dehydrogenous enzymes, uh, we go from having a small amount of

current that you see there in pink to a little bit larger in blue. Um, but what you expect to see is kind of a doubling, tripling, quadrupling of the current um, and the power that you get out of the um, electronic device. Uh, and it's not until we actually have all of the enzymes of the fuel cell, we vent um CO2 that we see this sort of large um increase in current density um and power density. And that's because any metabolic pathway, any natural metabolic pathway is going to have metabolic control within it. And so if we're actually not putting uh the uh all the enzymes down to be able to completely oxidize, we start to build up intermediates. those intermediates actually inhibit other enzymes of the the pathway um and and result in us having not as much uh sort of current and power um as we might want um to be able to power whatever device we're powering um with a fuel cell. We could

take the same strategy and look at other fuels. Uh and so if we go to the next slide, uh this is an example with lactate where we just took the exact same KB cycle um cascade, we added lactate to it. So we added one more enzyme um and you can see we get sort of similar current densities and similar power densities. And you might say well why is this important? Uh and basically if you think about uh an electronic device uh electronic devices uh need to be able to convert energy quickly in some applications. Um so if you're simply sending a text message uh that uh is just text uh no pictures uh no videos uh you can get away with a low power or a low current um fuel cell. But if you're trying to FaceTime, uh you need to um be able to convert those electrons um as convert that fuel to electric electricity as quickly as possible. Uh another example of this uh would be if you think about a transportation application, uh you can

have a fuel cell that may be able to go zero to 60 miles an hour in an hour and a half. Uh is not what you want. you want to be able to go zero to 60 miles an hour in several seconds and not sort of extended periods of time. So, so that speed of energy conversion or what is the the power um density of the device is important. The other thing that's important uh is actually the energy density of the device. Um and so if we look at the two examples I just showed you, I showed you pyuvate and lactate. Uh pyuvate uh is not very soluble in water. Um so if we um take it all the way to its solubility and we only have pyrovate dehydrogenase one enzyme, we will make a battery with about 70 watt hours per liter of energy density where a typical um sort of lithiumion battery would be in the kind of 250 um to 300 watt hours per liter. If we get do complete oxidation and we get every electron out of pyuvate um then we're

able to get something that's sort of comparable to the lithiumion battery. But if we look at pyuvate, the second fuel that I showed you, it's much more soluble. Um so because it's much more soluble, we can pack much more energy um into space into the same um sort of volume. Uh and so even with just a single enzyme, we can get 500 watt hours per liter um out of lactate as a fuel and we can store um 3,000 watt hours per liter uh out of complete complete oxidation. And I think this is important to point out because obviously we need high power in the form of being able to convert that um chemical energy to electrical energy quickly, but we also need to be able to store a lot of power so that we don't have to be recharging our um electronics all the time. Um so from that perspective uh I I think it's important um uh to think about the fact um that the uh the cascades of enzymes or using biocatalysis with um enzyme

cascades gives us the ability to both increase the power and current that we get out of a system but also the energy density. So that was just an example on the energy um side. Um but roughly uh seven or eight years ago uh I kind of expanded uh into um organic synthesis. Uh so at the time um Phil Baron and I uh had both just started as associate editors um for Jax. We met at a meeting um and he started um quizzing me uh about uh sort of electrochemistry uh with his focus on kind of electrifying um organic synthesis and I realized that this might be a new opportunity for us to go into new areas. Um so as an electrochemist I was really uh interested in sort of whether or not we could start a new um NSF center for chemical innovation. Um and so uh as we started to think about um basically building a center that would be trying to make uh organic electrochemistry or organic electroynthesis mainstream to

the organic chemistry um community. Uh, as an analytical chemist, I needed to go out and talk to the organic chemists and kind of figure out what their um pain points are, what their challenges were, and why they were still um not using electrochemistry. Uh, so I went out um to both academic and industry industrial um chemists and asked them, you know, why is it that you're using stochometric amounts of oxidizing agents or reducing agents and not using electricity or electrons? Uh as an electric chemist, it's obvious to me um I would never use stochometric amounts of oxidizing agents or reducing agents. Um but that's not so uh I guess obvious within the organic um chemistry community. So I started asking them sort of what their um issues were.

Um and the same couple of things kept popping up over and over again. The first was that they didn't really have any formal training. They didn't know how to use a potentiat. They didn't own a potentiat. And so that's a perfect thing um for a uh center for chemical innovation to address. Uh that's a workforce development um issue. Um and so we have been working on you know developing um curricula um to try to bring uh electroynthesis uh into the uh into the undergraduate um curriculum so that people feel more in their comfort zone. Um and from an instrument um perspective um Phil's group uh as well as the the rest of the center were alpha testing and beta testing uh the electroin at the exact same time that I was interviewing all these people. So um so I knew that was coming down the pike um and was um sort of less concerned uh about addressing the the no formal training issue. Um but the other issue

that kept coming up over and over and over again is that electrochemistry is not known for its selectivity. Um so it is known for having a pretty narrow potential window and not being able to get selectivity. And so people kept you know especially in industry comparing electrochemistry to photored redux and biocatalysis um and talking about sort of you know uh the the challenges um that that electrochemistry had and the one that just kept coming over and over and over again was you know if you're thinking about bio catalyst they have this great selectivity and electrochemistry can never get to that selectivity. Um, and at that point, as a person who's combined um, biocatalysis and electrochemistry for years and years and years, I I thought maybe this was really an opportunity for us. And so, I'm going to tell you a couple stories there um, in bioelectricalis and using enzymes um, as catalysts for organic

synthesis. Um, one of those uh, is kind of a focus on using the enzyme to give you um, uh, to give you selectivity. Um but uh sort of rather than having to add stochometric amounts of uh oxidizing agents or reducing agents which are typically co-actors in the biocatalysis um community, how we could eliminate um the need for co-actors and the other is kind of addressing some solubility issues uh that that come up a lot um when we're talking about organic electroynthesis with enzymes. So the first story I'm going to tell you is about CH activation of inert alkanes.

Um, as you know, we're still a petroleum economy and so a lot of our feed stock um chemicals uh a lot of our precursors are coming um from the petroleum industry coming from inert alkanes and needing to be able to activate um those inert alkanes um selectively. Um and obviously there's transition metal catalyst. Uh there are heterogeneous catalysts. There's a variety of um catalyst out there um to do that. Um but if you look at over the last uh decade or so, there's been a lot of work uh in sort of using biocatalysis uh and its ability um to have selectivity uh or to have selectivity engineered like the cytochrome P450 example um that's shown here um to be able to kind of um functionalize that inert alkan um selectively. And so um from that perspective you can think about a whole wide variety of bio catalysts that might help to to address that. But most of those bio catalysts are going to be

extored reductases. They're going to require um some type of co-actor to function. Um and co-actor uh regeneration uh can be a a challenging situation um to do in an economical and a green um fashion. Um so the story I'm going to tell you today is not about cytochrome P450. Uh it's actually about alkan monooxygenase. Uh the reason for this uh is that uh cytochrome P450s are relatively um easy to interface with electrodes uh but uh sort of non-hem containing um metallo enzymes like this alkan mono oxygenase are much more um difficult. So it was kind of a challenging electrochemistry um problem uh as well as an organic synthesis um problem. So, so we started with this alkan monoo oxygenase. Uh, we like it uh not just because um it's challenging from being a non-heem um iron oxidto reductase um enzyme um but also because it has a broad substrate scope. So it doesn't just do um hydroxilations, it does

epoxidations, it does suloxidations and it does demethylations. Um so it has quite a bit of um activity from that perspective. So the first thing that you have to do um in any of these sort of electroynthesis applications is actually get a hold of the proteins uh and we actually uh expressed isolated and purified all the proteins uh that are biologically responsible um for for doing this. So that's the alkane monoo oxygenase a mediator protein um called alga and putito reductinto redu reductant reductase uh that is doing your um co-actor regeneration. So expressed isolated and purified them did enzyme assays to make sure that they were still active um and then went um to actually see if we could put them into an electrochemical system where we did electroynthesis. The electroynthesis was actually um to drive the system without having to add any co-actor. No NAD or NADPH um needed uh and therefore not

having to add stochometric amounts of NAD or NADPH um NADH or NADPH uh nor um sort of adding any co-actor regeneration system. So essenti essentially what we're doing uh is basically removing biology's co-actor regeneration system and um feeding those electrons directly in from the electrode from the cathode um instead of uh sort of through a co-actor mediation system. It turns out this is a complicated system um and because it has this mediator protein um alga um and if that alga um sits on a bare carbon electrode it will um denature. we actually had to um sort of uh add a uh an electrocatalyst or a mediator um this tuine blue to the system. Uh this is electropolymerizable um so that we can get a nice coating on our carbon electrode um to keep uh that algae uh mediating uh and at that point uh we could use this cathode uh to actually look um at a variety of different um reactions. So this is a

monooxygenase. It's oxygen dependent. So we don't really see much uh electrocatalysis or electroynthesis until we add both uh the alkane in this case octane um and the oxygen. Um but when we do that we can actually look um at um the fact that it can do that hydroxilation. It actually um can do um our um epoxidations uh suloxidations uh demethylation. all of the chemistry. You keep all the selectivity and all the chemistry that the enzyme can do, but we're driving it um with electrons um from essentially renewable electricity rather than having to drive it with co-actor and co-actor regeneration system systems. Uh we can do that using a single enzyme. We can add enzyme cascades here so that we can do kind of one pot electroynthesis of more value added products um like amines. Um so we definitely have the possibility um to work there but we have challenges with this system um in that uh some of these

enzymes are oxygen um dependent and we have to actually control in that pot the time um where we have a lot of oxygen and the time that we have no oxygen. The other story I want to um tell you uh is just a really brief story. Uh the example is the electroynthesis of chyro um beta beta hydroxy nitrials. Uh another system um where you really want to get the selectivity um out of the enzyme. But unfortunately beta hydroxy nitrials their um reactants the intermediates and the final um nitrial um products are not very soluble. Um and so solubility becomes a huge issue with biocatalysis. Um that's handled a lot uh outside of electrochemistry with co-solvents. co-olvents actually don't work um in electroynthesis. Um and so we went to these bifphasic systems where we have a second phase that's acting um as a substrate reservoir and a product sync. Um and these systems do work. So we actually can um sort of increase um

the production of our product um by actually having a bifphasic system where we have an organic phase and a water phase. we have to make sure that that organic phase keeps the enzyme active uh while also having a good partition number um for the reactant um the product and the substrate. But if we do that uh you can see that we have uh significantly more um product um that's formed. Um so we keep that same um selectivity but we're able to produce significantly more products because we are able to improve the solubility. So hopefully in the last 22 minutes uh I showed you that combining pi uh combining biocatalysis uh with electrochemistry or what we call bioelectric catalysis does have applications not just to glucose uh sensors um that you might be uh um more familiar with um from going um to uh CVS uh or Walgreens and buying glucose test strips u for for diabetic um patients.

It also can have applications to energy conversion and electroynthesis. But there are still things we have to think about. Um we have to think about the oxygen sensitivity of our enzymes. We have to think about immobilization, enzyme immobilization and stabilization um for a lot of these applications in electroynthesis. We have to think about solubility and production rates. Um so it is a bit more complicated. So with that um I want to take just one last slide um to acknowledge uh the NSF center for synthetic organic electrochemistry got me into the uh organic synthesis um world um and uh has been great fun um to see how we can use our bio catalyst with electroynthesis uh as well as my um research group um past and present uh who have done um all the work that I showed you um today. And with that, uh, I will, um, turn it back over to the moderators.

Great, Shelly. Um, thank you so much for this fascinating and very insightful um, uh, presentation about um, using uh, uh, biocalysis and um, bio electrocalysis. Um, which also segus very well into um, Dooming's talk. Um um so um now I'd like to introduce Professor Dumin Drew uh very briefly um so that I can leave more time for his excellent talk. Uh Lumin has a long and balanced background between athemia and industry. He is currently the uh director of Tenjin engineering um research center of bioc catalytic technology affiliated with the tenjin institute of um uh industrial biotechnology of the Chinese academy of sciences and um his research entails the catalytic use of enzatic and biological tools to achieve formation of challenging chemical bond and productions of industrially relevant molecules uh from which field Duming will now present talk about his um bio biocatalytic synthesis today. So please

professor Drew. Okay. Thank you for introduction. See, so first I will introduce a little bit about what we are doing at the tip. So we did enzyme engineering and enzyme discovery and then we study the uh re reaction mechanisms and develop new synthetical methodologies uh for chemical synthesis. So after that we develop a process with the industry and collab collaborate with them commercialize the technology. So we study different uh enameric reactions. So today I will talk about two topics. One is engineering of DH GHDP dependent enzymes for the synthesis of high value added chemicals from formalhide and other alhides.

The second property is bio transformation of phyto stores to key intermediates for the production of st apis. As we know the utilization of carbon dioxide is a hot topic right now. uh carbon dio carbon dioxide can be first reduced to formic acid and then formalhide and then methanol. We are interested in the utilization of formalhide. So we are interested in uh the covering of artifice and to synthesize alpha hydroxy ketones and then transform this ketone to different high value added chemicals

for the coupling of carbonia compounds. Uh there are two types of reactions. One is form alpha hydroxy ketones. Another is form beta hydroxy ketones. uh the these reactions proceed through different reaction mechanisms and also catalyzed by different enzymes. So there are some report about the coupling of different uh adhesives with formal alihide for the coupling of formalhide with other hardihides. There are some challenges. The first one is a low chemical selectivity. From the reaction mechanism we can see from intermediate three when it react with the another alihide.

This alihide can be either formal alihide or the other alihide. So can form two different products also. uh this kind of enzyme show low tolerance of artifide because the artifide functional group can react with amino group in the protein and deactivate the enzyme also in some cases also show the low activity and limited substrate uh scope. To do this, we first collect a library of uh THDP dependent enzymes. Right now, we have more than 200 white type enzymes. We studied the catalytic properties of this these enzymes.

For example, we found this pip showed a relatively broad substrate scope for the artifact artifacts and area artifact adhesides. These enzyme show high activity for the coupling with formal alihide. But for sub seven and eight that's benzo arhite and benzo acet height this shows a low activity

also for this uh three hydroxy propanal coupling with formal adihide We cannot find any enzyme shows activity for this reactions. So we start with HIL. This enzyme shows uh good utraction level also shows a relatively good subst with enzyme. We uh first find mutant one shows a little bit activity and then so several round of mutenesis and screening we got a few uh bars shows a good activities

here shows the kinetic parameters of these mutants. But when we carry out the reactions use these mutants, we found M4 uh showed better performance. That's because M4 uh was more stable shows higher TM and we can we test at a different subject concentration we can say M4 shows better performance uh up to 300 mill more the conversion reached uh 95%

For the reaction of form alihide with benzo height the shows low activities also show low chem selectivity. One way uh use this uh reaction screening our library. We found uh this mutant enzyme shows a high activity for this reactions but but also shows uh low chem selectivity because the product 1B was formed. So we here we select the M3 as a starter enzyme to do uh direct revolution.

So we found a three mutants uh shows uh better cheml selectivity. We got the coal structure of these mutants. We found the movement of L 552 uh decreased the volume of the subject binding site also uh reduced the sub entrance entrance channel. So we can see this compared to the M3 uh_3 this uh sub engine channel is smaller than this one.

We also calculated the volume of the subject binding site. We can see the mutant M3 score and score 3 shows small much smaller volume. So this enzyme this mutant shows better chemical selectivity here from the reaction mechanism. We uh we uh can use the distance from this uh carbon carbon to this uh co-actor carbon 2 as the measure for the activity. So we can say for the mutant M33 most of the distance is less than 4.5 mm strong.

So this mutant shows higher activity and a better uh camera selectivity. So we use this mutant uh synthesize some uh alpha hydroxin meas. and cables. Uh this shows the three different mutants use the fed batch uh reaction mode. So we can see uh the S subj concentration can be increased to 700 mill.

Uh after we got this THDP enzyme and can make different alpha hydroxy hydroxy ketones. So then we focus on the transfer this alpha hydroxy ketone to different value added compounds. So first we synthesize this one to amino ac. So the immune reductor can catalyze the reduct reductive amination of different amines with different ketones. Uh synthesize these uh one two amino aros.

This is uh another another examples. This is uh at the preparative scale. Also we can uh combine these two reactions in one port two-step uh process to make this one two amino ac uh 29 and so two by chemical reactions and then one chemical reaction we can make this important compound.

We also use M4 uh combine this three reaction in one port. So uh one we can make this one two four D trail. Also we can combine this with conserminates to make this amino ac we can combine uh this carbon carbon uh form for formation reaction with ketone reduction reaction. reactions in one part uh concurrent um process. So these two reactions can uh carried out in one step. So we can make this one two diodes. This is the S steel isomer. When we choose another carbon reductase we can get the R as initial.

So if we carry out this reaction use a continuous flow addition method. So we can increase this substrate concentration to 900 m. So if we replace the carbonial reductase with amin dehydrogenase we can make this one two amino alose.

Yeah, this is shows this enzyme also can uh suitable for make alifhatic beta amino ac So we also uh combine this TH GP dependent enzy reaction with uh oxidation reaction and then uh reductive amination reaction to synthesize amino acids. So here use a oxidase to oxidize this amino this alpha group to the carbonic acid group to synthesize alpha keto acids and then transform this alpha keto acids to alpha amino acids.

We also got the mutant to uh catalyze the coupling of three formalize molecules to form once three dihydroxy acetone and then covered is a transaminous catalyze reaction to make this uh amino ac. So starting with a simple arose we can first make the hydroxy ketones and then transform this compound to different value added compounds like amino ac and also amino acids. So next I will talk about the by transformation of phyto stress to the intermediate for drug synthesis. So drug is a a class of very important medicines in the 12 200 small molecule drugs.

There are 12 uh steroid drugs more than uh 8 billion annual sale. When we start this project in China, uh we use the D gen as a starting material to make uh store APIs for the production of uh one pound this Jing. We need 100 tons of DS coal and also produce a lot of waste water from disgening to this intermediate also produce a lot of waste water. use a uh toxic chine uh oxidants. So we developed a uh process to transform phytos uh intermediates. The fossils is a byproduct from vegetable oil industry. So it's a uh it's a red abundant uh resource also is bio renewable based on the literature also our own studies on the metabolism mechanism of phytous by microacterian strains.

So we construct different uh strains. We uh for the synthesis of PA, AD, ADD and nine hydroxy AD. uh for n hydroxy add this compound uh this inter uh is not stable. It decomposed to uh this compound hip. This this uh deco composed to carbon dioxide and water. So we identified the the key enzymes and then constructed two strings to make uh hip and also hil.

After we got this string, we can uh did the transformation. We can see uh from original stream uh we got the C product is a mixture of different uh compounds. So for the new strings we can uh get this ad 9 O H A B A HIPP and H I L the content of this compound in the code extract from the fermentation uh mixture can be greater than 90%. So then it's easy for the downstream uh process. So we collaborate with a pharmaceutical company commercialize this technology. So we can right now we can produce these six intermediates at a time scale.

So right now in China for the synthesis of ST APIs is gradually uh uh started from the phyto stus to replace uh this genine. So this slider shows the product line. It has a company. So So this is the this stuff this production line is automatics. So it can be controlled in the central control room. Uh this because this is the first uh uh product line built by this company uh can be uh automatically controlled. So they built the nice uh central control room. So we you can see uh this production process on the big screen.

So right now we can use this production line to production to protect this six intermediates just use different strain because the starting material process uh are almost same We also uh construct con constructed two strings to make a compound 2E and 3D that the common 2E uh can be used to synthesize this uh API the hydroester uh this this API is very difficult to make because this 10 alpha methyl group for the natural sterile compound that 10 the me group at 10 is a 10 beta. So it's difficult to uh make this 10 alpha methyl group compound.

Uh this intermediates provide a new uh method to make this uh very important APIs API. Uh finally I will thank my collaborator and my team and also thank the funding from the national uh science foundation and also the fun founding from the industry partners. Thank you for your attention.

All right. Uh, thank you, Professor Drew, um, for this wonderful talk. I already have a ton of questions listed up um to, um, ask you a little bit later. And, um, I encourage everyone to also use the Q&A to write your own questions um, at any time. And u, we'll get to them at the end of the um, of the end of the talks. Um, and now I would like to move on to Professor Himing Chao. Uh from Ding's last slides, we can already see that automated production will be very important for large scale synthesis and I'm glad to say that Queen will be able to immediately tell us more about such automation effort. Um professor Jao also has an extensive prior experience in the biotech industry before returning to um academia and now at the University of Illinois at Urbana Champagne. He is currently the um Steve L Steven L Miller chair of chemical and biomolecular engineering and the professor of chemistry, biochemistry and

bioengineering. Uh when he has led an illustrous career in synthetic biology and incorporation of machine learning and laboratory automation and indeed today um he will give a lecture on expanding directed evolution to AIdriven automation. Uh the floor is yours uh professor Job.

Okay. Thanks for your nice introduction and it's my great pleasure to attend this webinar and for this audience I probably don't need too much introduction. Enzymes and microbial cell factories have been increasingly used to produce chemicals and fuels from renewable plant biomass and this actually is a sustainable alternative to the chemical refinery that we are all familiar with. However, compared to chemical catalysis, enzyme catalysis still have many, you know, limitations. Um, particularly, you know, they are not very stable and also not compatible with organic solvent. But the biggest limitation is really is the, you know, the reactivity is kind of very narrow, right? The type of reactions that enzyme can catalyze are still very limited compared to the reactions that catalyzed by uh chemical catalyst. On the other hand, the enzyme clearly has some advantages such as the high selectivity, environmental friendliness

and also com they are very easy to engineer compared to you know organic thingsis of the um liant used in organo metalysis and there are two actually big challenges in biocalysis. One is how to efficiently create enzymes with improved or new function particularly new reactivity and that is kind of the bigger challenge biggest challenge I would say and then second uh challenge is how to efficiently combine enzyme catalysis with chemical catalysis because I don't think enzyme catalysis will ever replace a chemical catalysis these are two kind of viable options you know the question is that how can we really take advantage of both type of catalysis this to create more efficient synthetic routes.

So direct evolution is one you know effective approach to address the challenges I just mentioned and this is near and dear to my heart because my PhD thesis was focused uh direct evolution. So I developed a number of uh direct evolution tools for enzyme engineering which also helped my former advisor Francis Arnold to get a Nobel prize in 2018. And the concept of direct evolution is very simple. Essentially starting from the parent gene, we just use a random gen searching recombination to create a library of variants and then use high through screening or selection method to identify the improved variance and then we'll repeat uh this uh and use the improve the variance as the starting point for next round of improvement and then we can repeat it again again and this is very effective for improving existing protein functions. However, it also has a major limitation. Uh it's a very hard

it's it's very difficult to use direct evolution to create enzymes with new functions or new reactivity because if you can just do a simple, you know, calculation, right? Uh the creation of new functions or uh reactivity typically requires multiple simultaneous mutations. So if you want to generate a library with three mutations, the possible library size is already more than 10 to the 10th which exceeded any kind of a selection or screening method that uh um we can develop. So clearly we cannot explore all those combinations to can identify new functions. So a few years ago not a few years ago actually more than 20 years ago we came up a strategy called in vitro co-evolution. So essentially we use some intermediate functions to bridge the uh gap between the y type of function and the uh novel function that we want to create. And then we use direct evolution to improve the function of those uh inter uh intermediate uh

intermediate uh functions and then we can gradually bridge the gap. In that case study, we use the liant bind domain of estrogen receptor as a test case and we try to create a new uh the liant bind domain that can recognize a completely new liant uh in this case is cautical and if you look at the chemical structure of cauticalone and the uh original sub original liant estradile they differ in many places right and the protein didn't show any activity towards the cortical serum but they did show some activity towards the intermediate uh compounds. So we can gradually improve the uh binding of the ligan binding domain towards those intermediate. Eventually we can create a new uh liant specificity and this strategy actually was used by Merc and Codaxis to engineer trans aminase with a new subicity which was incorporated in the synthesis of Geneva which it um anti-diabetes which is a drug used for treatment of

diabetes right so that is kind of a successful application of this strategy in industry And but this strategy is still kind of uh has many limitations because the process is still kind of uh tedious and also in many cases we don't know you know what kind of intermediate functions we should use you know in order to bridge the gap between the novel function and the uh wild type function. So about 10 years ago we came up another new strategy is to combine photocatalysis with enzyme catalysis. Uh because the photocatalysis is very versatile. It involves free radicals. So you can react almost all kinds of functional groups but it doesn't have high selectivity. While for enzyme catalysis we know it has a very high selectivity but it doesn't have broader reactivity. Right?

But if we can use light to create the radicals in the co-actors that enzyme use and then we use the active site of the enzyme to control the selectivity so that we can achieve both high uh selectivity and broad uh reactivity. Right? So that is kind of the simple idea and actually it worked very well. You know we initially demonstrated that uh we could repurpose the inred reductase that use F A and as a co-actor and uh we can generate the the radical on that uh FAD you know co-actor and then uh catalyze use the uh this artificial photo enzyme to catalyze the coupling reaction between the olifen and and this carbonial uh substrate.

Right? So as I shown uh uh here we can do the uh intermolecular radical hydroaculation reaction and also uh uh achieve the intermolecular radical conjugate addition and more recently we show we can uh introduce the fine motifs into the target product. All these reactions involved the carbon carbon coupling reaction and um the product we can produce all highly inential uh pure right so that's something that you cannot uh achieve uh using chemical catalysis uh readily but for enzymes we can do that and also in addition to the carbon carbon coupling reaction we can catalyze the carbon nitrogen coupling reaction as well to uh achieve the radical called hydro amination reaction and all these reactions actually follow the same kind of a mechanism. So we basically use the light to create the radical on the co-actor and then use the light to create the radical in one of the substrate right and then the radical

substrate can react with the olifan substrate to form this prochyro compound and then the enzyme active site controls the hydrogen atom transfer in an enantial pure uh enantio selective manner to achieve the high selectivity And here I just want to show the uh florination reaction uh briefly. In nature there's only one class of enzyme so called florinase that can catalyze direct florination and uh um but but the the substrity of the enzymes are very limited because they all use a kind of a same uh kind of like they all use the same as the substrate, right? So you really cannot change the substrate specificity that much and also we know that the uh incorporation of florine can really uh change the property of the uh molecules right almost 30% of the pharmaceuticals or the agricultural chemicals have florinate florine in it.

Yeah. And so there's a a huge demand you know uh um to create new like a florinated uh chemistry right and our idea is to uh as I mentioned in the previous slides we actually uh add a leaving group um uh to the chlorine and then created the uh radical and then we uh introduced another olifant substrate and then we actually used the enzyme to catalyze this coupling reaction on the light and we can create a wider variety of uh products with uh florine that shows high selectivity and this reaction is kind of unusual. So we tried many different florinated uh agent and we found that the try florine io iodine iodo ethin and this uh florine agent works well at least it can achieve high you know selectivity but the yield is very low and also we made sure that this is a enzyatic reaction right it doesn't you know it's not like a spontaneous reaction so we did some control studies to make sure it is

indeed uh enzyatic reaction and we also tried many you know different substrate to show the broader scope but the uh we also uh explored this reaction at a different remote site we can achieve uh the selectivity at the beta or gamma or even delta position so at the delta position I'm not aware of any chemical catalyst can you know achieve selectivity. But with the enzymes because the binding pocket is pretty big, we could actually uh you know catalyze the inential the the stereo uh to achieve the selectivity at the uh remote uh site.

But for uh in order to do that I think the white type of enzyme is also not very uh effect uh effective. So we also have to use direct evolution to further improve the uh the selectivity right. So we have to introduce mutations and then you know find uh uh look for the mutants that can improve the uh selectivity and in addition we did some computational studies as well try to really understand what determines the uh selectivity right why it produce uh why the enzymes has a preference for S product rather than R product and also we identify the residues that are responsible for the stereochemical uh control.

So photo enzymes clearly you know they are very versatile. They can catalyze many different type of reactions but none of them is really used in industry right at least right now. I think one major limitation is that the calic efficiency of these enzymes are very low and um uh so you have to use lots of enzymes in the uh reaction which made the whole process not econ economical. So that's why in the past uh year we try to come up a new strategy to address this uh uh challenge and our idea is to incorporate the artificial photo enzyme in cellular metabolism and we want uh um use uh the whole cells right uh to produce the two uh substrates used uh by the photo enzyme and also uh uh use glucose as the uh feeder stock right the substrate because the glucose is very cheap you know compared to any chemicals we use in catalysis.

So uh as a proof of uh uh concept we uh engineered a cell to produce this uh olifant right and then also produce this uh uh substrate and then we have an engineered photo enzyme that can catalyze this coupling reaction on the light to produce uh uh this uh uh bioactive compound. And so we did a lot of uh enzyme engineering because um even um we incorporate the photo enzyme into cell metabolism its activity is still very low. So we uh did a direct evolution uh study to improve uh the uh overall activity of the enzyme and also did a pathway engineering to you know generate the the two co uh two substrates uh and then we uh have uh this engineer the cell or you can think of this more like a programmable catalyst right it is basically we program everything. We programmed the substrate, programmed the catalyst, and then we only use the glucose as the feeder stock to produce the catalyst and substrate and product.

And when we when we did the uh fermentation study, we found that it can produce uh like a 6 gram per liter. And this is not a very high tighter, but at least it shows uh a demonstrate the proof of concept, right? It is possible to incorporate engineer the photo enzymes into cellular metabolism and this will open up uh the uh product uh a lot of opportunities right basically now enables to produce many compounds that cannot be produced uh using natural uh enzymes or even you know chemically. And now I want to switch the gear. I want to talk about how actually uh uh we can use uh AI and robotics to accelerate the enzyme engineering effort and I already explained the principle of direct evolution. But the new paradigm we try to pursue or establish is so-called autonomous experimentation. So in this paradigm we use AI tools like like generative AI tools or large language models to make predictions on

the enzymes that we should create. Then we use robotics to perform all the experiments and then you also use the supervised machine learning model to predict the combination of mutation that can further improve the protein function and then we can do it irreratively. So the principle is similar to direct evolution except that we use um AI to make a predictions and guide the experience and also all the experience will be done by robotics rather than a human researcher and I think the power of both direct evolution and AI driven autonomous expression depends on the iterative cycle right is we have to accumulate those mutations very quickly and I started this effort actually more than 10 years ago go. At that time I was inspired by development of computer on our campus. Um more than 70 years ago, a group of physical uh a fittist um on our campus designed one of the first computers called Illinois automatic

computer one or illac and it consists almost 3,000 vacuum tube weighted about five tons, you know, but the computing power of that computer already exceeded all the computers um uh uh in the bell labs combined. And so about 10 years ago we built the uh uh integrate robotic system called the Illinois biological foundry for domestic bio manufacturing or IBO fab. So just like a computers for computing, we used the uh the IR5 or bio funy for rapid prototyping of biological systems. And this uh effort led to the formation of global bio alliance in 2019 which currently consists of more than 40 members around the world and also many actually in industry and also led to the formation of the NSF bio funy last year which I will talk about in a minute. So our idea is to really integrate a lot of uh instruments into a large robotic system as shown in the video. You know the robotic arm will take 96 well plate you know from one

instrument to another right and uh there are more than like 20 instruments uh on the platform. So you can think of this is a automated lab and we use uh uh this lab to kind of perform all kinds of uh synthetic biology uh experiments. So we break those uh complex workflows for protein engineering, metabolic engineering or just library screening into process modules which will be in turn broken down into unit operations and then we all program those unit operations to create customd designed workflows for different uh uh applications and uh particularly with the incorporation of AI we could actually achieve autonomous experimentation and we studied this effort in uh 2013 with the uh grants from uh Cava Trust and then you know obtained the grants from various funding agencies including DAPA, NIH, NSF, DOE um so basically uh further developed the uh technologies for the bio funary but also explore the

applications of bioundry. So in 2020 we built the second generation of bofundry and then just uh a month ago we had a ribbon cutting for our third generation uh bio funary called I bio funary one with the funding from uh NSF. So we used uh this bofundry to achieve autonomous experimentation um a few uh years ago and this is a proof of concept study. We try to use multiple enzymes to synthesize an important chemical called lycopine and we um uh express those enzymes in the whole cells. But we need a balanced expression of these enzymes to achieve high uh yield. Uh so what we did is that u we use uh um the robotic system to construct a very small library of pathway variants that contain uh promoters with different strength for the three structural genes and then we use the robotics to perform all the experiments right and also quantify the product and then we use a very simple uh AI tool called a basian optimization to

um analyze the um uh data to make a predictions on the combination of mut uh combination of promoters that can further lead uh to high um uh that can improve the production of lycopine and then we can do it irritatively. So we only made 46 uh variants in each round, right? And basically here I showed the a workflow. Uh we constructed the uh plasmid that contains those pathway variants and then transformed them into E.coli for production of the lycopine and then we use the pl reader to quantify the lycopine production and then use the AI to to make predictions and then we can do itly. We repeated the three runs and in total we designed a construct uh like 138 um uh pathway varants which is less than you know 1% of all the total uh possible combinations right so the take-h home message is that with this AIdriven pathway optimization algorith uh par uh workflow we can significantly reduce the experimental effort you know

we can find the best optimized the u um pathway very quickly and a similar approach now has also been used for chemical synthesis or material discovery as well. So and this uh proof of concept study uh helped me to obtain a grant from NSF in 2020 to establish one of the first seven AI institute in the United States called a molecule make labor institute or MMI. So in this institute we try to develop a new AI, new chemistry and biochemistry to radically accelerate and democratize molecular synthesis and functional material discovery. Um so there are five thrusts in the uh uh institute. The first two all focus are the development of AI tools for synthesis planning uh catalyst selection and optimization. Uh so we are not limited to enzymes. We also look at chemical catalyst as well. And our idea is to combine you know chemical catalysis with uh enzyme catalysis right to achieve more efficient synthetic

routes. And then in thruster three and four we try to demonstrate those AI tools uh for different applications. In thruster three we choose three small molecules. Two of them are drugs. Uh the third one is a molecule used in polymer processing and in thruster 4 it is organo photo attack material. We try to improve the efficiency and stability and recently we just got the uh center renewed for another five years and so here are the key deliverables in the center. So basically we try to develop all the AI tools for molecular synthesis and we call it the alpha synthesis.

Basically the AI tools could be used to mine literature or existing database and also the data that we will generate using like bioary or similar you know uh systems for chemical uh uh synthesis as well. And then we have this database that contains the chemical reactions enzyatic reactions or the mutants you know for different proteins. And then we um leverage this database try to develop a new uh AI tools for synthesis planning, catalyst selection and optimization. And another key deliverable is the AI powered closed the loop for uh the engineering of catalyst. And here we uh used the four kind of case studies. Uh one is related to Suzuki coupling reaction. try to optimize the yield of the reaction and other two are related to enzyme engineering and also the fourth case studies organo photo attack materials. We try to improve uh the stability and efficiency.

So here I just highlight a few AI tools that we developed. Uh one is this AI tool for predicting uh enzyme function. Man is the enzyme commission number that will indicate what reaction the enzyme catalyze or what are the substrate it will use and many of you of course are familiar with this alpha fold uh which is used to predict the protein structure from sequence and that was developed by deep mind you know a few years ago and they also um now is widely used you know in protein science and engineering fields. uh while with our tool we now can really know the uh enzyatic reaction uh for the uh proteins and also more recently we developed another AI to to predict the enzyme subicity because even we know the enzyme reactivity like what the reaction to catalyze but still there are so many small molecule that could be used as a substrate so in many cases we don't know which one is the best so

that's why we also developed a follow-up tool to predict the uh enzyme substity. So we can rank you know the molecules uh using this AI2 and predict which one will be the best you know for the particular enzyme and in addition to those AI you know tools for predicting enzyme functions we also developed the AI tools for enzyme engineering as well. So uh almost five years ago we developed AI to to predict mutations that can improve protein function particularly it are those high order uh reaction uh mutations meaning that the combination of uh several mutations that can improve the protein function.

And also more recently we um combined the AI with the robotics to achieve autonomous uh enzyme engineering. So in this new paradigm we use the protein large language models to predict mutations that can improve protein function and then use the robotics the bioundry to perform all the experiments including site derto genesis characterization right and then based on the data we generated we use the AI model to predict the combinations mutation that can further improve the protein function and also we develop a chatbot you know to help uh the research just to design experiments as well. And as shown here, we used two enzymes as a test case. One is a methyl transferase that used as a a catalyst. The other one is a phase that is used as animal feed.

And uh so we performed four rounds of uh uh kind of a direct evolution. So in the first round we used the protein large language model to predict more like 180 mutants that can potentially you know improve the protein function and then we made all of them and quantify their activities and we found that indeed ranging from like 20% to 50% of the mutants uh showed the improvement. This is much higher success rate than random metagenesis right that people use in classical direct evolution experience and then uh and then in the second round we use a AI model to you know predict the combinations. So we construct a much smaller library of variance only like 96 of them and in the subsequent run we all construct no more than um uh 96 variants. So in total we uh constructed and quant uh characterized less than 500 mutants and yet we can significantly improve the enzyme function right as I shown uh here.

So now of of course we also try to make our technology you know broadly available. So with the funding from NSF uh last year we established this national infrastructure called IO bioundry in which we try to develop a cloud bio funy that seamlessly integrates synthetic biology AI and automation to democratize basic and applied biological research. So we have a inhouse uh research program in which we will continue to develop AI tools and automated workflows for enzyme engineering and cellular engineering but also we have a user program in which we want to engage the broader research community. So we will solicit the proposals and we will fund those proposals and encourage them to use our facility and all the tools that we develop or you know we could collaborate to develop new tools as well. And also we um launched another NSF center focusing on Bio Fundry but mainly through collaboration with um uh

partners in other countries as I mentioned in the beginning. Now there are more than 40 uh academic bio countries right in in the world. So we particularly partner with the bio fun, Japan, Korea and Finland. And our goal is to develop the standards and the matrix to make the bio funy more reliable and scalable. Uh currently you know bio funies all use different equipment, different workflow, different tools, different ways to manage data, right? So it's very hard to share the workflows. So we basically try to create some standards among the community. So in summary what I talk about today is that we developed a new strategy for enzyme catalysis by combining um uh uh photocatalysis with enzyme catalysis to create photoenzymes with new to nature reactivities. And I showed a few type of reactions that can be uh catalyzed by these new class of uh uh artificial enzymes. And also I I I talk about the

AI institute uh that I lead um to develop new AI tools for molecule discovery and synthesis and also I introduced a few AI tools for uh enzyme function prediction and enzyme engineering right and we're developing many other uh AI tools as well and particularly I think the take-h home message is really is the AIdriven autonomous experiment experimentation I believe is a new kind of a research paradigm and this also can be applied for chemical uh synthesis and material discovery and senses as well. So in uh uh finally I would like to thank the students and posters who did uh most of work. Many of them have left my group already but particularly I would like to thank DOE which provides most of funding for us to develop the capabilities for the bioundry and then uh NSF in recent years for the development of all the AI tools for um uh molecular synthesis and discovery. So, thank you for attention

and uh I'll be happy to take questions. Great. Um thank you to all our speakers. Um we have time for some Q&A. Um but I just want to start by thanking everybody for some really insightful and interesting talks. I think as someone with a background in organic chemistry, it's it's wonderful to see how the beautiful reactivity that's being obtained through biocatalysis. Uh you know, I come from a background where we're used to sort of engineering very clunky enzymes iteratively and the power of being able to just screen you know hundreds of enzymes for for a desired activity very rapidly is just it's fantastic. So we have a number of questions in the chat. Uh we are coming to the end of the allotted time. So we'll run for about 15 minutes in the Q&A. Um I I'll try and sort of synthesize the questions into broad themes uh so that as many people as possible can have their questions answered. Uh one of the themes that has

come up from a few people in the questions but for the panel as a whole is about scale about the challenges of of large scale. So for example Emily asked about if there are particular challenges in scaling uh bioelectric catalysis versus conventional biocatalysis. We have questions about the impact of AIdriven enzyme optimization on industrial scale biocatalysis. So I wonder if anybody from the panel uh could speak to uh you know particular challenges and opportunities uh in industrial collaboration and and large scale biocatalysis.

I think Emily's uh question is excellent. So if you you know look at the biocatalysis community there's a lot of effort in sort of bioreactor designs uh automation of of sort of you know bio foundaries um as as we saw um with humans talk and uh and so I think you know technology there is evolving um but at the same time we're just starting to use bioelectric catalysis for kind of electroynthesis applications um and so there aren't bioreactors that have really been designed to have electrodes in them. Um maybe electrodes from a sensor perspective, but not electrodes in terms of of of the um electroynthesis um that that would go on there. And so um so I think you know it's a great opportunity. We've been you know talking to a lot of people about sort of reactor um engineering um recently um because an electrochemical bio foundry would have to be um something different um and is definitely an area we're interested in

going in. Thanks. I I wonder if uh Professor Jur in particular, you mentioned that the institute you're at does a lot of collaboration uh with industrial partners. I mean, I wonder if you could speak generally about how that works and you know, if there's any advice you would give to researchers in academia in the audience who are interested in in working on this scale and getting uh collaborations with industry established. Is there are there any particular challenges whether they're technological or social or cultural that you've encountered there that you can speak to?

Yeah. Uh for for the collaboration with the industry uh uh I think the important thing is uh establishing uh trust between the uh collaborators. So uh also need uh use a proper uh mechanism to do that. So we usually establish long uh longtime uh collaboration with partners if we we do not we usually do not like only one transform. So we started from the get the u industry question together and then work together and then uh push this technology to the uh commercialization together. So it's kind of difficult. Yeah. It's a not like a publish a paper.

Yes. uh it's a whole other order of magnitude when publishing a paper. Um there is also quite a specific question here I think about the role of AI uh on industrial scale um reaction. So I'll read the question out because it's along this theme but I think it's quite timely. Uh so this is from um a researcher at Jangju University who asks what the key technical bottlenecks facing the application of AIdriven enzyme optimization in industrial scale biocatalytic reactions and what exist what effective breakthroughs have been formed in existing research. I mean this maybe a question for professors jao perhaps as you you've talked quite a lot about uh AI in your your talk and your work.

Yeah, I would say currently there's a renaissance in enzyme engineering. Uh really is because of the development of AI tools, right? Currently there are many you know startups now focusing uh using AIS for protein engineering and also you know in industry like the you know existing companies as well and I think the reason is because you know we have those protein large language models that really enable the prediction of protein functions and also engineering. So there are many tools now we can use to accelerate the enzyme engineering effort and particularly those robotics also help the uh actual experience greatly as well. So I think there's a um lot of there's a um a lot of activities you know in this field and uh I cannot point out any like real success yet uh but you know this field is also very new. I'm sure in the next few years there will be many kind of industrial example successful examples that really

use the AI tools for engineering and also scale up. Certainly as an editor, you know, we get a lot of research in the last five years using large language models and AI and various applications and I think enzyme engineering, protein design and also materials discovery are areas where it seems to be having a real impact and real success versus some other areas where maybe it's a little more uh hype driven. Um

I mean it's a very specific question for you professor Jiao uh about your AIdriven pathway optimization workflow. Um you've been asked uh how do you determine the appropriate number of design and test iterations and specifically when to decide to stop the cycle and then following on from that how do you verify the final results is the truly the best producer not a local optimum within the design space. Yeah, I mean this is a good question. This is also a question for any direct evolution experience, right? In reality is that we stop. So one the uh the students run out of time but now we have the robotics so we can continue in principle, right? But so then the the the the stop point is really that once we meet the performance target performance matrix, right? So we know that how you know much improvement we have to make in the activity of stability and then we just use the AIdriven autonomous expansion platform

do it and see that what uh whether we can meet on them or not but actually recently I I was asked what happen if we run the experiments like a thousand times or even more right what will happen because because I I I'm curious about that you know so we probably will do that in the near future scale is very impressive. Um while we're on the subject of of you know technical technical questions and specifics we have one I think the first question we got actually in the chat uh comes from Alice who has a broad methodological question which maybe um for anyone on the panel about how to identify effective high throughput workflows for identifying you know enzymes with the desired function and the particular she's particularly interested in thinking about how to prior prioritize candidates that can translate from say a bacterial model to a mamalia model. Is that something that anybody in the panel

can can speak to? Yeah, we are recently developing an enzyme retrieval AI too. So basically and this is a very good question because there are so many you know enzymes out there right how do you know the enzyme you use is the best out there right now we're not talking about the enzyme engineering because with engineering you can always improve the protein function but just you know uh pick the starting point right how can we find the best enzyme for any like experiment so I think you know we really try to develop some AI tools um to First one to identify the enzymes right for a given reaction. But then the more important thing is also like the enzyme property prediction tools we developed and we now can predict the the like a stability, the catalytic efficiency, the solubility all kinds of properties, right? And and then you can have a kind of a prioritized list of enzymes for testing and then that will help you to find the

best enzyme very quickly. Now those AI tool is not 100% accurate but it can at least help you to narrow down the targets and I think this kind of a strategy will be increasingly used in the future.

When you refer to AI models in this context are you referring to sort of just machine learning in general or are you specifically thinking about the current wave of of generative AI live language models? not a gener it's those AI tools that like I showed two examples right we developed the AI tools for predicting the EC number also the uh the the substituticity but there are also many AI tools for predicting the kat km and the solubility stability of the enzymes right so now you just need screen all the enzymes out there and then you can identify the one with the best you know properties out there and then you can use the other one to do further engineering or further testing, you know.

Thank you. And um this for the last couple of minutes we have uh one final question for professor J um about the construction of chyro centers and how your team balances precision of enzyme structure analysis with the efficiency of engineering strategies in the context of constructing chyal centers for a desired product. Oh, this is a yeah, this is a good question, but it's actually it's very hard to answer this question. Yeah. So usually we just like a fing. So we use uh computation to predict first predict and then um uh then use the experimental to verify and uh then do analysis again and just it asks uh do it several long and then get the part enzyme.

Well, I say I think we're a little bit over time, so I think we'll have to wrap up here, unfortunately. But thank you to all of our panelists for joining us and for sharing such exciting research. And thank you to everybody for listening and for sharing your questions. Uh there will be a recording of this uh email to all registrants on Monday and it will be online on cell.com next week. So if any of your colleagues missed out and they're really excited to uh follow up after the fact and hear about the exciting work we've discussed today, then they'll be able to do so. So again, thank you to all our our panelists and thank you everybody for attending.

Okay. Thank you very much. Thank you. Do the fight.

Questions about this episode

Who speaks in "Biocatalysis: Sustainable innovations in catalysis for a greener future"?

Shelley Mir (Main speaker (panelist)), Heming Chao (Main speaker (panelist)), Jeff Van Drew (Main speaker (panelist)), Shiao Shiao Chao (Host).

What is discussed in "Biocatalysis: Sustainable innovations in catalysis for a greener future"?

In order: 0:00 Shelley Mir, Jeff Van Drew, Shiao Shiao Chao; 10:00 JAX; 20:00 Jeff Van Drew, Chinese Academy of Sciences; 50:00 Heming Chao, National Science Foundation, University of Illinois at Urbana Champagne; 1:10:00 Lycopene, AI Institute, Illinois Biological Foundry, Global Bio Alliance; 1:20:00 AI Institute.

What is mentioned in this Cell Press episode?

15 people, works and ideas are mentioned, including Shelley Mir, Heming Chao, Jeff Van Drew, Shiao Shiao Chao, National Science Foundation, NSF Center for Synthetic Organic Electrochemistry. Each one on this page links to its own page with every episode that mentions it.

How long is "Biocatalysis: Sustainable innovations in catalysis for a greener future"?

The episode runs 1 h 38 min; the transcript is about 12,465 words in 98 timestamped sections.

More from Cell Press

About this transcript: the episode's spoken words as captioned on YouTube, split into timestamped sections. Reclip.Me locates every person, work and idea in the text, keeps the time it is first named, and quotes the words around it verbatim. Last updated January 23, 2026.