11. Protein Design and the Communal Brain w/ David Baker

In this episode, we interview Dr. David Baker, a professor at the University of Washington and the director of the Institute for Protein Design. We talk to David about the practical and philosophical sides of protein design, the impact of machine learning on protein structure prediction and design, and how video games can be used for science.

For more information about EBRC:

Visit our website at ebrc.org. If you are interested in getting involved with the EBRC Student and Postdoc Association, fill out a membership application for graduate students and postdocs or for undergraduates and join today!

Transcription:
Episode transcripts are the unedited output from Whisper and likely contain errors.

Transcript

0:00

Hello and welcome back to EBRC and Translation. We're a group of graduate students and postdocs working to bring you conversations with members of the engineering biology community. I'm Kevin Reid, a graduate student in the Alpert Lab at UT Austin. And I'm Andrew Hunt, a graduate student in the Jewett Lab at Northwestern University. Today we have the pleasure of interviewing Dr. David Baker, who is the director of the Institute for Protein Design, a Howard Hughes Medical Institute investigator, and a professor in biochemistry with adjunct appointments in many other departments at the University of Washington. David, we're really excited to have you today. Thanks for joining us. Great. Thank you. Looking forward to it. To get us started off, can you tell us a bit about your journey to becoming a professor and protein designer? It was very circuitous. I had no intention of becoming a scientist when I was an undergraduate in college. My initial major was social studies, then philosophy. And I kind of got into science in my last year. And then when I wrote my graduate school applications, I said I was interested in neurobiology and developmental biology. And my graduate work was on yeast cell biology. I never touched a computer. In fact, I ridiculed anybody in the lab who touched a computer as a total waste of time. And then I decided to make a big switch for my postdoc. And actually this is, I think, relevant to graduate students out there. There were a number of people who knew me as a graduate student who I remember running into people who said, oh, you had dinner last night. We talked about your insane postdoc choice because it was in a completely different area. But then the first day of my postdoc, there was this computer terminal on my desk. And I asked what it was for. And the PI said it was for computing. So I had to learn what that was. And so I guess the moral of the story is you can do lots of different things and career trajectories don't necessarily go in a straight line. And then towards the end of my postdoc, we got really excited about sort of the fundamental principles of protein folding. And that's what I applied for faculty jobs, saying I would study the simplest possible cases of protein folding and figure out how they worked. Very cool. So sort of building from that, you spent a lot of your early career focused on the structure prediction, but really have transitioned your lab largely to de novo protein design or protein design from scratch. Can you talk about what drew you to de novo protein design initially and why you're still passionate about it now, maybe 20 years later? Yeah. Well, let's see. Like I said, when I first started in Seattle a long time ago now, we really focused on experiments, trying to figure out how proteins folded. And then sort of as an adjunct to that, we developed computational models for protein folding and found that they worked pretty well for protein structure prediction. And after we had gotten pretty good at protein structure prediction, we realized that we could invert the problem. Since we could map from sequence to structure, we realized we could go backwards from structure to sequence. And Brian Coleman came to my lab and really sort of pioneered that. And once we started being able to design new proteins, then this huge brand new world opened up of being able to make brand new proteins from scratch just for fun or to solve real world problems. And it was much more expansive, I would say, than protein structure prediction, where it was really protein structure predictions just sort of taking naturally occurring sequences and predicting their structures. Whereas with protein design, like I said, you could create this whole new world. So I think that was what drew me in. Yeah. It's neat as you think it's a little bit related to your initial interest in philosophy as an undergrad. Yeah, you know, it's funny, you know, you look back at things you've done. And I would say that all these things that I did early on, like what was what was the use of being a social studies or philosophy major? But now I and why and being a yeast cell biologist as a graduate student. But actually, I think all those experiences do sort of add up. I mean, what we do now is really, really broad. And so I think knowing a little bit about a lot of things and being able to think critically in a lot of different ways is important. Like one of the reasons actually, I got out of the, I sort of got into design away from protein folding, because at that time, there was just a lot of talk about how proteins folded. And it was kind of like, kind of almost sort of, it wasn't very rooted in reality. And that's a lot of what, you know, you sort of think about as, you know, as a philosophy student, do arguments have content or not? Or is it just sort of language games? And with protein design, it was very, very concrete, you know, you make your design and either it works or it doesn't. And then having a background in experimental biology was really good when we started to actually test designs, because, you know, whenever you make design, say a new enzyme, you really want it to work. So if you make a bunch of designs, it's easy to have kind of rosy colored glasses, you know, that make you think it's working when it's not. And so I think having some familiarity with how to do rigorous experimentation was really important. Yeah, that makes a lot of sense. Yeah, your comments on rooted in reality feeds well into the next question. So your lab has traditionally developed tools for protein structure, prediction and design based on first principles, like Rosetta. However, new approaches leveraging machine learning like alpha fold and Rosetta fold have received a lot of attention. Can you talk about what you see as maybe the utilities and limitations of each approach? Yes. So the deep learning approaches are extremely powerful when there are high quality and large data sets to train models on. And I think one of the most important conclusions that can be drawn from the success of alpha fold and Rosetta fold is that the protein structure database is now rich enough to contain really detailed information on sequence structure relationships down to, you know, the positions of individual atoms in a sequence of previously unknown structure. Where physical models are important is when you want to compute things where there isn't a lot of data to go on. For example, if you want to model unnatural amino acids or non-protein like entities, if you want to compute quantities like binding energies where there aren't large data sets. Now I have a little story here. I have a student who is constantly bemoaning the loss of, you know, the truth and beauty and the physical models for this kind of, you know, obscurism of the deep learning model. So what I say is, you know, that there may be a historical analogy, you know, in ancient Greece, you know, 500 BC, it was truth and beauty. Everything was principles and transparent and full of light. And in some ways, you know, high point of human civilization. And then, then we kind of went into later on, we went into the dark ages and it was just magic and mysticism. And but then at the end, then after that, the Renaissance came and it was like everything came together and the truth and beauty were even better than before. So we may just be in the dark ages now with the mysticism and magic of models with hundreds of millions of free parameters. Yeah, that's a lot of fun. Do you think that we will get to a spot like that where we'll be able to pull out physical insight to build better physical models? Yeah, I mean, I think it's a really exciting time. Everything is moving. Things are moving incredibly fast. And undoubtedly, there will be going back and forth between the physical models and the deep learning. So there are a couple examples. For example, right now, sort of quantum chemistry methods have really embraced deep learning. And so those methods do very, very well. We're starting to going the other way, we're starting to incorporate more physical modeling into Rosetta fold, and it's clearly making a difference. So I think it's a little bit, I wouldn't want to predict the future, but all I can say is going to be very exciting. Well, in the interest of predicting the future, our next question is about that, what do you think are some of the next major challenges in protein design and in structure prediction? Well, in protein design, there are some of the things I'm really excited about, as you know, developing really improved therapeutics that currently protein therapeutics, it's largely antibodies, which are kind of blunt instruments. They have a single combining site or bivalent that goes and knocks out a target. But for a lot of issues in medicine, would one want some more sophisticated medicine that could do logic calculations in the body, make decisions, and so forth. And then there's issues like, could we design proteins that cause cell fate transformation? So for example, turning cancer cells into differentiated tissue or for recovery from injury, promote regeneration. Outside of the biological realm, nature gives us many, many exciting examples, you know, tooth and bone and seashells. They're basically proteins mediating the deposition of inorganic compounds by mineralization. And I think there's just huge potential there because imagine all the things you could make if you could get, you know, say, semiconductor materials to form on a protein array, or even calcium carbonate, new types that you could, if you could make designer things like bone or tooth or shells, it'd be incredible. Light harvesting, you know, nature is really, really good at it. Catalysis, you know, there's a lot, there's a lot of bad things we put on the planet. So if we could make enzymes to break them down, that'd be fantastic. So there's really just a huge, I could go on and on, there's lots and lots of exciting applications. And what's really fun for me now is I get really brilliant people coming here, as you know, who are eager to make the world a better place by designing proteins to solve one of these problems. Yeah, totally. And all of those applications require, you know, expertise from many different areas, right? So one of the other things that we notice and that I noticed, especially when I'm reading your work is that all of your publications have contributions from multiple labs and across different disciplines. How do you manage that collaborative process? And what have you learned through collaborating with so many different people? Yeah, well, I think the key thing to solving hard problems is getting smart people together who are really excited about the problem, who have really a range of expertise that covers all sides of it. So for my graduate students in postdocs, if they're, when they're working on a problem, and I always say, well, find the best person in the world, or the person you think would be most exciting to work with, and let's contact them. And I think that's so important. My model for my group is sort of a communal brain where all of these, you know, sort of like you can have individual researchers kind of work on their own. And that's like, you know, many, many small sets of, you know, individual neurons. Whereas if you get everyone together and everyone's kind of talking and brainstorming together, you get this, I think really emergent properties happen, you can get really high order accomplishments. And I think basically collaborations are a great way to extend that to brilliant people who are experts in all these areas all over the world. So as far as managing it goes, you know, I think I send a lot of emails to people asking if they've been collaborating, and I actually get a lot also. So a good example of how important that can be is we were interested in using, well, we used Rosettafold to predict the structures of all the core complexes in eukaryotes by sort of pairing all pairs of proteins, then predicting how they interacted. And we were left with this huge amount of data, but it was very, very hard to get biological insight from them. So I spent three or four weeks just contacting people, and my students did as well, who had worked on these complexes. And it became a whole village trying to interpret these different things. And it was super fun. And we learned just so much. So yeah, I think collaboration is absolutely critical. And I think you just have to, you know, you just have to contact people. And usually most scientists are excited about collaborating. As far as managing them, my students and postdocs generally, you know, have direct interactions with the collaborators, which I think is really the best way. So I'm not a limitation. It's interesting to me, because you really get a lot of information about a lot of different disciplines, right? You're crossing traditional protein biochemistry with now lots of engineering biology stuff. But also, I mean, as you mentioned, there's enzymology and biomaterials and all of this stuff. Can you talk a little bit about how you, I don't know, synthesize all that information? One thing is I just, you know, I always like new things. I get kind of bored easily. So just having some new area to go into is just super exciting. That's one of the really fun things. And, you know, I'm really fortunate that, like I said, these brilliant people come in who have the domain expertise I don't have. And so I can often learn from them, or else we'll have a really close collaborator who we work with very closely who provides that. So I don't ever pretend to become an expert on anything. That's fair. So I think sort of along these lines, you also manage a really big lab. You have like almost 100 or maybe even more than 100 active members at really varying different points in their careers and from diverse backgrounds. Can you talk a little bit about how you lead such a large team and mentor a really diverse group of people? Yeah. So like I said, the first is the communal brain principle. So one of the big things is creating an environment where everyone's interacting with everybody and brainstorming. So there's no hierarchy. So it's totally flat. So everyone's kind of equal. Everyone's a creative and contributing ideas to the group. So we have two group meetings a week where three people each present. We have happy hours after each other. So it's a lot of social engineering just to get people talking all the time. And I don't travel and I don't do very much teaching and I'm on no committees. So I basically spend 100% of my time when I'm not doing things like this, just talking to people in the lab about the research. I meet with every student or postdoc every three weeks and most of what I'm trying to do is connect people. And then I guess the most important thing is that people come here are really brilliant and excited and enthusiastic. So there are all these spontaneous collaborations bubble up, which are usually the most exciting ones. So we've also heard rumors that you actually do have projects that you work on yourself. Could you talk about that a little bit? Yeah. Well, you know, that's been a casualty of, well, it's indirect casualty of the pandemic. So up until the start of, you know, the lockdown, I always had my own project and I never really had a schedule. I just sort of wandered around the lab every day, trying to find somebody who wanted to talk to me. And if no one did, I worked on my own project. But with the pandemic, since we were doing everything over Zoom, I had to have a schedule. So, you know, I had half an hour blocks all day talking to students on Zoom. And then once the pandemic ended, that has stayed. So right now I'm a little embarrassed to say I have not made much progress on anything on my own right now. So what type of project would you work on? Well, I think my most recent projects were deep learning projects because it was a couple of years ago we were trying to do deep learning and I wanted to learn some. So one of my most recent projects was to see whether you could compute the energy of a protein structure from just the backbone coordinates. And that is useful because in Rosetta, when you do that, it's very time consuming because you have to do this search over all the side chain confirmations. So it didn't work very well, but I learned a lot. I'm curious, are you planning to pick any of those back up once things, you know, maybe go back to normal? Yeah, I'm still trying to figure that out right now. I think I would have to go back to my less scheduled life to do that. And yeah, I think, you know, everything's, I think we're all still in flux on how our lives are going to be once things get totally back to normal. So I think I'm still figuring it out. Well, that's great. Switching gears to something else, something really neat that has come out of your lab is the Citizen Science Focus Project, Foldit, where your lab created a video game for players to predict protein structures and design new proteins. Can you talk a little bit about how that originated and how it's grown since its inception? Yeah. Well, part of it is, as I described, you know, I've always been interested in sort of recruiting excited, smart people to help solve hard problems. And so this is kind of Foldit is an effort to really go wide on that. The way it started is before Foldit, we had started a project called Rosetta at Home, which is still ongoing project. And that provides a lot of the compute power for the research here. And so Rosetta at Home contributors, they contribute spare cycles on their computers when they're not otherwise using them. And that's really what's powered a lot of the advances in de novo design over the years. So we had people, when you run Rosetta at Home, you see a screen saver appears that shows the course of your protein folding up or being designed. And people started writing in saying, well, I'm watching the computer do this. And I think I could do a better job because the computer just kind of randomly searching. And then I like going to mountains. I decided to go up for a day with our went for a day with the father, a friend of my daughter's who was a computer scientist. And I was describing this. And then we started talking about, well, maybe you could have an online video game, which was like Rosetta at Home with Rosetta underneath where people could go in and start modifying things. And then he introduced me to a couple of really talented graduate students and a CS professor interested in games. And we were kind of off and running. Really cool. Can you talk a little bit about what it's used for today? Because you all still run Foldit, right? Yeah. You talk a little bit more about like, if I were to log on to Foldit right now, what sort of stuff would I get to be involved in? Well, let's see, there are different types of challenges. One is designing protein assemblies. It's kind of fun in Foldit. So it's now set up so that you can choose a symmetry, like say cyclic symmetry, sign with five copies. And as you change one copy, all the other copies change in synchrony. So it's kind of like looking through a kaleidoscope or something. So that's fun. Actually, Foldit players were had puzzles trying to link the different coronavirus binding domains and we're still working on, you know, the problems in proteins, small molecule recognition, protein DNA design. So the problems, the puzzles change every week or two, and they usually reflect some of the main challenges that people are interested in here. And we put those puzzles on Foldit to see what people come up with. Oh, very cool. In addition to all of your academic work, you all have spun out a really large number of companies over the years. Could you talk a little bit about the entrepreneurial process from the academic point of view and maybe about when you think a technology or a discovery is ready to go build a company around? Yeah. Yeah, it's funny, because the last thing I ever thought I would be doing would be spitting out companies. Yeah, that just shows that's another example of don't plan too far ahead. Yeah. So the origin of that is after we started doing protein design and started making proteins that could be useful, the students and postdocs working on those projects got interested in sort of continuing them, but we didn't really have funds for that. So we were able to raise some funding for what we call translational investigators. So if you come to the IPD as a graduate student or postdoc, and then if you create something that looks like it could be useful, the program can support you for another year or two of getting whatever it is you've designed to the point where you can launch a successful company. So that relates to the question of when is the right time. I think that the most successful companies I've been a part of are ones where you really have something that already has value. And it just, say in the case of a vaccine or a drug, there's already some demonstrated efficacy and it's really a matter of planning how to get it out in the real world. And I think the nice thing about our translational program is you have the time and the resources to actually develop things to that point. But now it's kind of cascading or accelerating because now many people are coming to the IPD with the hope of then starting their own company when they move on. And so I think we're in the process of starting three companies right now. And, you know, they're in all of this, they're covered with this really wide range of areas that we work in. So yeah, it's a really exciting time. It's fun for me because it's great. So then, you know, normally when a graduate student or postdoc leaves, then they go on and they do something completely different. But now people are really pursuing their vision. And, you know, I get to, it's just really cool to see all these different efforts going forward and not just ending. I mean, that's the traditional criticism of academic research is it doesn't really impact the world. But now having sort of this route for people to go out and further develop what they've done as as graduate students or postdocs is great. Yeah, that's great to hear. And I'm curious, could you talk a little bit more about what your what your role is in these companies? Is it mostly just purely advisory or do you do any like, do you have any additional responsibilities? Yeah, well, let's see, my role is just advisory. But when people leave my lab, it's a little bit like kids leaving home. When they ask me for advice, they kind of know what I'm going to say. And they're probably tired of listening to me anyway. So so I would say that I not asked for advice all that much, which is fine. You know, I think I I provide whatever, you know, I'm very supportive. But and I try and do anything that is would be useful so I can help make connections and stuff. And yeah, but I you know, my my focus is really on stuff going on at the IPD. Yeah, that's great. What did what advice would you give to someone who's really interested in the field of protein design and maybe interested in becoming a professor? Well, I think if you want to become a professor, I think one of the things is just to remember that's your vision and don't lose sight of it. And I think determination and motivation is really important for that. Obviously, you know, doing really exciting, great science is is the best recipe for success. I would say be an environment where you're kind of set up to succeed. And so that's kind of like with my group, I sort of see that as my role is like creating an environment with, you know, the people, the brain power and all the other resources that anyone would need if they really wanted to do exceptional science. It's you know, it's hard to do that in isolation. So you want to be in a place where where everything sort of set up for you to succeed. That's great. And it sounds like you I think you had mentioned previously that you don't do any teaching and you're not in committee. So you sort of carved your own path as being very specific research type professor. How did you negotiate that? Or was it just over time, you sort of evolved into that? Yeah, well, let's see, I told you I didn't really have a calendar up until recently. Well, early on in the department, I think it was realized I didn't have a calendar. And so I was I have this bad habit of missing meetings and stuff. And I was always very nice about very apologetic, but I think I just wasn't very good on committees and stuff. So and then I have I do teach a graduate biophysics course every fall. So that that was just sort of what the job came with. And I think now at this point, I think I can I can just be very direct, I can be very open to someone. If I get an email asking me to give a talk somewhere, I said, Well, I'm, you know, I'm really focused on research here. So and talking to students and postdocs my lab, so I can't travel, but I can do it by zoom. And so I can just say that. And then I think, you know, people here know that I'm busy, and that I'm really focused entirely on research. So, you know, you kind of develop your style, and then people get to know you. And, and then it kind of all works. That's interesting. Yeah. So another quick follow up question. So I really liked that there's this article by Paul Graham that has like the the maker versus manager schedule. And so I'm curious how you balance like you thinking about ideas versus you like, talking to students and like, you know, in postdocs and guiding them on their own projects. I kind of think, you know, I think they're sort of the same thing. Because a lot of the ideas I get, and I think a lot of the ideas anyone gets, and certainly people here come up from just conversation. So I think it stimulates thinking. So I sort of look at so when I'm meeting with students, it's there, it's usually sort of each, there's usually brainstorming, you know, so I mean, I do have ideas when I'm walking to work, or, you know, we're skiing on the weekends. So but I, I think that that that constant interaction is sort of an idea generator. And that's why I when one of the one thing I do if I if a new student comes into the group, and I don't see them talking to a lot of people, I, you know, I really encourage them to just just go and talk to as many people as they can about what what they're doing. And yeah, we had a group meeting last week, where two groups of three people presented on projects, super cool new projects they had done. None of the three people had any their research had anything in common before they were just sort of ideas that kind of bubbled up. And yeah, it was really beautiful. So I kind of think that talking about the manager, I wouldn't say I don't really feel like a manager. I feel like, you know, these conversations are about science, and we're kind of thinking about ideas. And I think that's that is partially where good ideas come from. So as a last question, we were wondering, in an alternate reality, where you didn't end up going into science and didn't end up becoming a professor, what do you think you would have done instead? Gosh, I always like new things. And and I get kind of, yeah, I don't know, something where I could where things were moving quickly forward. I think that's what really got me from philosophy into science is that that, you know, philosophy and social studies were just they were static, they weren't moving, and then science was moving forward. So there's a sense of progress. I don't know. I mean, I, I, I'm not really sure it's it's I mean, I guess I probably would have been happy doing a lot of different things. But as I mean, I can think of a lot of fun things. I mean, I love climbing mountains and, you know, exploring the world. But I don't know, I would have I'm sure I would have enjoyed being a 15th or 16th century explorers, explorers, you know, trying to map out continents and things. But today's world, I don't know. I mean, yeah, I I'm curious a little bit about the the mountain climbing real quick, because I'm like, my family's from Colorado. So I do a lot of like the 14ers and do a lot of mountain climbing like that. What what are your what are your mountains of choice? Well, I actually was funny. I like I said, I don't travel very much. But I got invited some some years ago by the University of Colorado Boulder graduate students to give a seminar. I said I would do it if they if we could ski down a Forkiner. I don't remember which one it was. But that was really fun. I like you know, this time of year, I like doing a lot of ski mountaineering in the spring. And yeah, the cascades, of course, Northwest is great for that. So awesome. Yeah, that's exciting. Well, David, is there anything you'd like to promote for being on the podcast? Things like DEI efforts or research openings or papers or anything like that? Let's see. Well, there's a lot of things, but I don't want to go on and on. And I'm worried if I give you a list of a couple of them, then the ones I like leave out. I think I'll just maybe leave it there. Yeah. Well, thanks so much for coming on the podcast, David. It's been a pleasure talking with you. Yeah, great. Well, that was a lot of fun. Thank you. So this has been another episode of EBRC in Translation, a production of the Engineering Biology Research Consortium's Student and Postdoc Association. For more information about EBRC, visit our website at EBRC.org. If you are a student or a postdoc and are interested in getting involved with the EBRC Student and Postdoc Association, you can find our membership application linked in the episode description. A big thank you to the entire EBRC SPA podcast team, Catherine Brink, Fatima Anam, Andrew Hunt, Kevin Reed, Ross Jones, Kogsi Lee, and David Mai. Thanks also to EBRC for their support and to you, our listeners, for tuning in. We look forward to sharing our next episode with you soon.