If computational chemistry sounds intimidating or unapproachable, this interview with Dr. Gabe Gomes will change your mind. Gabe is super personable and talks about how his team is applying computers to automate parts of chemistry while still letting humans shine where we do what we do best. Join us to hear how chemistry will be done in the future and how we’ll get there!
Visit https://www.thermofisher.com/chemistry-podcast/ to access the extended video version of this episode and the guest content sheet, which contains links to recent publications and additional content recommendations for our guest. You can also access the extended video version of this episode via our YouTube channel to hear, and see, more of the conversation!
Visit https://thermofisher.com/bctl and use the code Chem2Life to register for your free Bringing Chemistry to Life T-shirt during March 2023.
This started with a TV in the background showing Brazil playing Croatia in the World Cup quarter-finals, and ended with Brazil’s surprising defeat, to the dismay of our guest, Brazil-born Gabe Gomes. In the middle, the most approachable conversation you’ll ever hear about computational chemistry.
Gabe tries to solve real world problems using computers and it’s almost a paradox that such an extroverted, fun guy, in love with music and speaking so much about people, ends up investing his life in machine learning algorithms. Yet it takes courage, creativity, and daring to go in new directions and seek the next big problem at the interface of scientific disciplines.
Chemistry is a complex multivariate problem and resolving this complexity is the key to the fundamental understanding we need to advance the discipline. Gabe is a wonderful chaperone in our journey to discover how automation and optimization can be used not to replace chemists, but to free them to apply their skills where in matters most. Gabe is the living demonstration that computers and humans can be part of the same discourse.
Dr. Gabe Gomes 00:06
Science is made by people, right? Like there's no way around it. So you we should not try to not bring our whole selves into the conversation.
Paolo 00:17
Dr. Gabe Gomez is a scientist and an accomplished one with that. But he's also a musician. If you ask me, he's a bit of a philosopher too. And by bringing all of these interests and curiosities to his work in catalyst design, his science is as humanly intriguing, as it is scientifically normative. In the season four episode of Bringing Chemistry to Life, we speak with a member of Chemical and Engineering News 2022 Talented 12, about their work and trends in their fields. I'm your host, Paolo Braiuca, Director of Global Market Development, Thermo Fisher Scientific. We began by asking Gabe about his upbringing, Brazil, and his early opportunities in science.
Dr. Gabe Gomes 01:00
I really come from this really, really small town in the countryside of Rio de Janiero state. You know, back then, when I was playing guitar in a band, and also doing a chemistry technician course, during my high school. I had the opportunity to go to the Federal University of Rio de Janeiro for a weeklong series of mini courses, where you learn, you know, all sorts of things. And it was really cool, because it was the first time that I saw a university, but also, first time seeing people that looked like you know, like, me and regular Joe, roaming the, those spaces. Before then, you would I think that like, “Oh, these scientists are Schrödinger and Einstein and whatnot.” And you'll see these people and you're like, “Wow, that's interesting.” And there's that, but that was complex, because no one in my family had gone to college. So it was not really like something that we talked too much about. But I kept asking questions that my professors in this technician course would say, “You know, you'll learn that in college.” I was like, okay, I went to college and first week, there was this guy, who comes in puts Schrödinger’s equation on the board. It was like, second day of college. And he's like, “All the chemistry is in this equation. All you need to do is solve it.” And it all of my naivete, and I was like, “Wow, that's amazing. How does that work?” And like, little did I know that it was way more complex than that. But his course was all about trying to understand what happens in molecular geometries, molecules, in different geometries will impact their energies and reactivity. At that first semester, this guy, his name is Pierre Esteves. And I started to work with him, and in that year doing computational chemistry, and he was a person that was traveled. And he had liked this big way of inspiring us to ask deep questions. And also just like, “Careful what you dream about, because it might become true,” he would say that. During my undergrad, I just started doing mostly computational chemistry, but also some experimental chemistry. And then what happened was, there was a student from Florida State University and her P.I. that she was presenting, they did a tour in Brazil, talking about Florida State University. And they had this program called LASER, Latin American Scholars Education Research, something like that, that they would pick about four people from Latin America to go to Florida State's chemistry department for two months, and they would fully fund this this person to do to do research. I went there and worked. It was great. And then I went back for my PhD. But at the end of my, of my PhD, I was seeing what was going on with computer science, specifically for machine learning, and I started to get really interested in the idea of how to augment chemistry with machine learning.
Paolo 04:27
First of all, it seems to me that you're a sort of growth mindset, right? And in most cases, it's a natural thing you either have it or you don't and it's hard to develop right? There are people who like to stay in their comfort zone and people who get nervous in their comfort zone they want to push out. You’re the latter for sure.
Dr. Gabe Gomes 04:45
You are right. Another thing that I really like is, I like to go and see, for example, what people in design are thinking, right? I really try to bring in design thinking into my way of thinking about research. I like to see what people in art are thinking. How can we think about this in terms of policy? How is this going to be?
Paolo 05:12
I like you're going there because that would have been my other question. You know, you, you, you mentioned a few times as you were describing your path, you know, the sort of multidisciplinarity, right, of your experiences. And that was what actually fascinated me of your work. I mean, I look at some of your papers and stuff is, you know, how too often in my experience, people focusing on computational chemistry or whatever flavors, they tend to kind of get fixated with problems that become a bit sterile, right? It feels like something that you keep in mind all the time, even for the more theoretical of your works, right? It's this connection to chemistry. And so the sort of bigger questions you're trying to give an answer.
Dr. Gabe Gomes 05:56
This didn't come to me naturally. That part, it was very much a part of me of my mentors, how they think about it. Both my undergrad and my PhD mentors really think about, you know, if we're doing computational work, first of all, we must be rigorous and really use the tools to the capabilities and capacity that it can take us, right? We should really not try to get answers, that the technique is not performant enough, let's put it this way, to to give us that answer. And I also think that I almost don't like the division between computational and experimental. To me, it's almost irrelevant, because quantum mechanics works extremely well. But the reality is that chemistry is still an experimental science. And those two things come together, which is incredible to me, to be a professor at the house of one of, you know, one of the greatest pillars in computational chemistry. We had here, John Pople, that spent a good chunk of his career here and got a Nobel Prize in Chemistry for well, developing model chemistries, right. And I think that if we look at John Pople's work, although incredibly rigorous, and mathematical, it, you almost my opinion of course, can feel that, like, there was this, this thinking of, we need this for, to answer questions in the real world. And this is something that I really, really, really try to pass on to my students, as well. It’s like, “You know, it's great that we do all of this. But if it's something that you cannot get in a way that our colleagues, experimentalists, cannot understand it's pointless.” We need to have this deep integration, and really think about how these things talk to each other. And the other day, to me, and to a good chunk of my group and research program, there is no difference, I've really don’t say, oh, this is cooptation all night. But you know, like, when you're when you're applying for things, and submitting things...
Paolo 08:26
Well, of course, there's some rigors because you need to fall in some buckets. And we all understand that. It's yeah, but it's, you know, what, what does it really mean? You know, why am I doing this? You know, what kind of, what kind of answer am I giving? Right? is it like, am I just generating, you know, an image in a computer, which is very rigorous, and, you know, and very precise, or is it, am I trying to get an answer that somebody can take and apply somewhere, you know, am I advancing the field in any sort of concrete way? That's what, you know, everybody should think about when they do chemistry or science in general, regardless, as you say, we shouldn't even speak about theoretical, computational, or you know, wet chemistry or whatever it is, it's all the same science, isn't it? Just looking at it from different angles.
Dr. Gabe Gomes 09:17
And I'll say, I'll say, I'll add one more twist on this, which is during my undergrad, the program that I did have a lot of chance of industrial chemistry and subjects of chemical engineering. And I am now a professor in the departments of chemistry and chemical engineering. And the whole idea of learning from different angles is incredible, right? Because like now, like, I am working with these people that have a very different view of the world, but it's really, really amazing. And we have these exchanges where students coming into my group that have more chemical engineering background, they have a view that is so amazing for many of the things that we want to do at the same time, they mix up with the students that come from more chemistry background and learn from them. And they learn from my colleagues as well. So I don't know, like divisions. I personally don't like labels, don't like divisions. And I think that this boils down to.
Paolo 10:19
Diversity is richness, right? And this is the way you describe it, if you if you are good enough to get inspiration from different flavors of the same field or from different fields altogether. You know, that's where the biggest leaps usually come from. There's a couple of, you know, concrete examples of your work that show this sort of multidisciplinarity, right, the way you connect the computational chemistry ideas with, you know, the sort of wet chemistry concepts. You know, but it would be nice if you could start by describing, you know, your work in some way. And perhaps we can pick one example, I have the one of you know, the cross Kaplan optimization that you presented at your Talented 12 event, that was an interesting one to me, because that was a bit of a connection. But if you have another example, I'd be happy to go wherever you like, really. But you know, let's start with you know, the description of your work from you. I mean, that will be better than if I try.
Dr. Gabe Gomes 11:17
So, I like to think that we're trying to solve problems in chemistry, chemical sciences, and engineering using computers. And I like to think that we are interested in discovering and optimizing new chemical reactions, new catalysts. After all, we all like to talk about, you know, we had this amazing information revolution over the past 10, 15 years with the advent of smartphones and so on. But have you ever thought about the amazing technologies that we've had for 200 years of manipulating atoms, right? That is what we do as chemical scientists. And specifically when it comes to catalysts, it's one of the most powerful engineering tools that we have. It's the tool that allows for us to develop the food to feed the ever growing population. It's the tool that allows for us to have the medications that saves us from pandemics. It's the tools that allow for us to be having these very podcast right now with all the silicon technologies that we have. And all of that comes from developing new ways of breaking and forming bonds. So that's how I think about our research. So when it comes to examples, like the one you were talking about, so that was a collaboration with Merk, where the idea was there is a complex cross coupling reaction that had been trying to be optimized by hand using, you know, the regular techniques that have been used for many years that have been successful, like design of experiments. But they're still having a hard time. And what was very nice was that Merk has these high throughput robots, that they could do reactions in high throughput. So we sought to use machine learning technique that comes, that falls into the class of active learning called Bayesian optimization, where we then parameterize this chemical space in a way that could be represented in in these probabilistic models. And then the model goes and learns that hyper surface and basically tells us, “Okay, the next turn should be with these variables, so that you can learn more or maximize that objective function.” And that's how we are tackling problems now. You know. We, in my group, as well, we're developing Bayesian optimization tools, for example, that incorporate mechanistic information about a given reaction so that you need fewer and fewer data points to optimize it. Or we're developing inspired by nature and inspired by Pittsburgh native Professor Francis Arnold of Caltech, that developed directed evolution, we have a program called computational directed evolution that we are using for development for developing and optimizing catalysts substrates to understand how certain enzymes do the reactions that we want them to do. And so it's really this blend. And we also have the experimental lab that we're at this point we are building but it will be a lab that we won't be doing things by hand. It's only with robots that are controlled by either a Bayesian optimizer or reinforcement learning approach where we have this deep integration between chemistry automation and the realization in real space.
Paolo 15:18
There's a lot there, right, there's so basically the way you're describing, you know, chemical problems are in actual effect. So highly multivariate situations, right? Highly multivariate problems that you're trying to solve, in some ways do you need by a standard design of experiment approach becomes tricky, once the number of variables in the control and you have over them is over the, you know, the number of variables is exponentially growing. And the control you have over them is actually quite limited in many cases, right? So, and you also combine experimental and more computational sort of data together. So, using certain novel approaches in machine learning, you can actually try and mix a more diverse and heterogeneous type of variables, and probably being able to handle much bigger ones and limit the number of experiments you actually need to run. Right? Is that a sort of a good summary of this is the direction you're trying to take here?
Dr. Gabe Gomes 16:19
Yeah, it is. I will say, though, that we do have interests that are more fundamental, as well, that, you know, for learning how certain external effects can impact a catalyst, for example. But we are really interested in these questions of how we can accelerate reaction discovery and optimization. And, you know, we are, at the same time that we're doing all of that it's true, two points that I'm very, very happy. So, I am incredibly lucky to be part of two NSF centers. A phase two Center for Computer Assisted Synthesis, that, you know, it comes with some of my, honestly, my heroes like Abigail Doyle, Matt Sigman, Richmond Sarpong, Hosea Nelson. And also a phase one center, which is the Center for Chemoenzymatic Synthesis, where we're trying to learn how to control enzymes for very specific reactions that we'd like to do. Learning what are the smallest pieces of information that we can gather and make it to the smallest catalysts possible from those systems. So this is the work that we're doing with the incomparable Alison Narayan at University of Michigan. And as well as Scott Miller, Joe Hardy, this is the kind of thing that I am super super enthusiastic and happy to be working on. But then, at the same time, I'm bringing, being brought by my engineering colleagues, my chemical engineer colleagues, because it's into things like developing computer vision approaches to be able to count viruses in in a sample viral capsule, and capsids, in order to develop viral vaccines, for example, right. So like, we are really trying to do things that are quite on the edge, and it's super exciting.
Paolo 18:41
We hope you're enjoying this episode of Bringing Chemistry to Life. Stay tuned at the end of the episode for information on how to access content recommendations from our guests, as well as information on how to register for a free Bringing Chemistry to Life t-shirt. And now back to our conversation.
Paolo 18:59
I have a couple of questions, you know, when for what I understand, I've read about, you know, this this, this is a growing, right sort of automation in chemistry, the use of robotics, you know, it's an interesting expanding field. It seems to me that people so far have been mostly focusing on the reaction that can be somehow compared to many compartments with things since they're very repeatable, you know, that you can standardize in some ways. So it looks like if you look at the chemistry, you run their way and you look at you know, the chemistry that are normal synthetic organic chemists would run routinely, they look, they look kind of different. And I guess there's you have to deal with the limitations right of the automation itself, or maybe because we are at the start of a field or perhaps because we need to think about chemistry in a different way. Perhaps there will be as you automate these things, and then you start using machine learning. Perhaps chemistry will look different in a few years from now. What is the way to kind of interpret this in your opinion?
Dr. Gabe Gomes 20:02
Yeah. So something that I think all the time is that there is no way of, there's no way of not having very specialized skills and hardware to perform the kind of reactions that you as you're describing that are mostly done by, you know, highly skilled organic chemists, for example. And that's okay, that's okay. Because there's a lot of things that we can automate that will allow these highly skilled organic chemists to do more complex problems. Right? So that's fine. But what I do believe strongly is that I don't think that we should be trying to automate things in the same way that we do things by hand nowadays. That is the part that doesn't make much sense to me. Because if we're going to go through the trouble of automating something, we should make the best out of the tools that we have, instead of liking how we do things now. So I think that the way that we should be thinking more and more and more about is the way that Kerry Gilmore at the University of Connecticut has been thinking about if he's designed for radial flow synthesizers. Kerry and I are friends and collaborators. And that's the kind of hardware that we're going to have here that Kerry has many there. Where it's a very, it's a different way of thinking how you do the chemistry, flow chemistry is not a brand new. Kerry didn't invent this, but Kerry invented this particular design that allows for much more complex chemical connections.
Paolo 21:51
Definitely, well, for chemistry offers some advantages, right? For hazardous reactions, it's kind of easier to scale up in some ways, but you know, very often, it needs a lot of catalysis in general, right, it needs a lot of optimization to really perform the way it should. It's not something that you plug and play.
Dr. Gabe Gomes 22:09
Hence why Bayesian optimization and we’re working together...
Paolo 22:11
Yeah, and so the way the way you're describing is not that, you know, the future is going to be robots, replacing humans in labs doing chemistry. It’s that you can actually combine it to, there are problems that you can take without automation. And there are things that you know, are skilled, organic chemists will still be probably the best way to go, at least in the medium term, isn't it?
Dr. Gabe Gomes 22:31
Absolutely. And in fact, that is the vision that we have here at CMU, the automation, physical sciences. CMU, will inaugurate next year, the very first academic cloud lab in the world. It's a forty-five-million-dollar investment where you submit your experimental set of experiments it's called and it will be performed by a mix of people and machines. And you know, it this democratizes the science. Imagine that, you know, you can't be in the lab for so many hours, because maybe you have to take care of a loved one, or maybe you unfortunately, cannot operate in that lab. Now you can still perform experiments. So it's this mix of people and machines performing these experiments. And this is a community in collaboration with a company called Emerald Cloud Labs that is, was formed by two CMU alum. And they are already operational in California. With this, this schema of, you know, you submit your experiments and it's performed by a mix of people and machines. And I think that this idea will become larger and larger. So, CMU is investing in this, so that we can expand who will have access to performing these experiments, or any experiments that they would like, right?
Paolo 24:04
The whole idea is the work that you're doing, and you know, you know, the people you collaborate with and in general in the field, and with results like you know, that we are having with you know, in the protein folding prediction, for example, you know. All these are mind blowing things, do you think? Do you think chemistry is at the verge of a revolution.You know, is chemistry going to look significantly different from you know, in 5, 10 years’ time something?
Dr. Gabe Gomes 24:31
I have no doubt, I don't know if we're going to see it as something like in one day woke up and it's a completely different it's not like that right? But certainly in five to ten years, it will be much more digital for sure. And much more, computers will play a much bigger role. And you know, we see things are fantastic, like, alpha folding that you're that you're describing, and you know, more recently ESM fold by the folks at Atlas. We are really, at least in that part, it is, it is revolutionary no doubt about it. We are using those tools in our group for research that a year ago or two years ago, I would never even think that we would be able to tackle. That's amazing. Now imagine, you know what, what else will come in the next few years. Even with what we have now, with these large language models, you may have seen ChatGPT, that has been going the rounds on social media. It's a really amazing and fun tool. But we must be careful, because a lot of these models give you something with such confidence that you may not even think is right or not. Right? So that's where skilled and well-trained scientists would have a role is to be able to discern fact from fiction. That's something that I am curious to see how we are going to...
Paolo 26:19
We'll see. Now 10 years, we are young enough. We'll still be around. Anyway, I'm the way you're describing here, obviously you like the challenge you like the problem, you're, you know, you're probably a problem solver, right? In your anatomy. I'm just curious what takes you out of bed? Is it the challenge itself is you know, the sorts of knowledge that you can crack a very difficult problem, or is a is it you know, knowing that what your solution might potentially mean, you know, to the real world, what is what is really your ultimate driver?
Dr. Gabe Gomes 26:31
It has changed as I've been changing. Like me, as I've been aging, and also my role has changed. Before it was like this immense drive was knowing what is you know, the results, I'll be able to get that I could be the first person in the world to learn about, you know, something, and there's nothing like that. And you know, that remains, but now it's a little bit different, because it's more about training these people that will be on the driver's seat on that and learning from them and seeing them grow. It's just so, it's blissful. When you see them have these step functions in learning and growing it and thinking and cracking those problems, it's just so rewarding to train the next generation. But at the same time, you know, we have to, we have to act fast, right, we have to be able to develop a lot of these tools in order to guarantee a cleaner, more sustainable future for the generations that will come after us. And that is something that I do think a lot about. And sometimes I get very nervous because it's like, okay, we'd not being fast enough. But there's this immense sense of, training these folks at the same time learning what, you know, what challenges are they able to solve? And how can our solutions and ideas help to accelerate how we discovered materials that will keep us from climate catastrophe?
Paolo 28:43
It's great. I mean, it's great, any, I don't think we speak enough about that, you know, how important you know, the role of people like you is for the real world, you know, because you can crack problems that could actually make them work better. But also for the lives of so many people that you know, you kind of meet and work with you or for you, you know, and you train and educate. Yeah, it's amazing. Yeah. I love that you actually rationalize that in spite of what you are.
Dr. Gabe Gomes 29:10
Yeah, that let me just correct one little thing that you said, that “work for me.” No one works for me. We work together.
Paolo 29:22
Thanks. That makes it much better. Yes, you're absolutely right.
Dr. Gabe Gomes 29:26
It's a small thing. But it's a thing that it's not just because that's the right thing to do. It should be. But also, and this is important in the training process that people feel the responsibility for the problems they're trying to solve and ownership about it. Right and it's not you're working for me. It we're working together.
Paolo 29:56
Working together on something. Yeah. Going back to what you were saying at the beginning, you know, it's sort of the team spirit, right, this chemistry between people, and it's so important to you. It's great to hear. And that's a perfect segue to my final question, which is always the same. Yeah, now, you know, now that you know, you've grown up, you're still young, but you are also in a position to be able to, you know, support others, right. So what would be the single piece of advice you'd give to someone who's just starting in their career?
Dr. Gabe Gomes 30:32
First and foremost, keep an open mind. Try to learn from others from other fields, from your colleagues from, from literature, from not just, you know, try to really see what other areas might need that you can bring to them. So this open mindedness, I think it's important. But you know, I also think that whatever you do, try to be kind to others and to yourself. In our field, in academia, and science in general, it's very easy to fall into traps that relate to impostor syndrome, or you know, relentless attrition. And I just don't think that that's a healthy way of trying to do things and, and this is perhaps for myself, too, you know, like I'm saying this, but, you know, impostor syndrome is something that I have every day. And I still think that I am not good enough to be here and shouldn't be here. And kindness to self, I think it's something that you should have. And the last thing I'll say is like, yeah, seeing what's going on now. It's great, but you should also be trying to see what's coming next. And that's hard. That's just hard. But I would say like, you know, try to see what's coming next and that open-mindedness will lead you to it.
Paolo 32:14
That was Gabe Gomez, Assistant Professor of Chemistry and Chemical Engineering at Carnegie Mellon University, and one of the Chemical and Engineering News' Talented 12. Thanks for joining us for this season four episode of Bringing Chemistry to Life and keep an ear out for more. If you enjoy this conversation, you're sure to enjoy Dr. Gomes' book, video, podcasts, and other content recommendations. Look in the Episode Notes for a URL where you can access these recommendations and also register for free Bringing Chemistry to Life t-shirt. This episode was produced by Sarah Briganti, Matt Ferris, and Matthew Stock.