In this episode, I'm attempting to string together the most common threads brought up by my guests in the first eight interviews on the podcast. Reflecting on these conversations, I'm suggesting some grand challenges for the field in the years to come.
In this episode, I'm attempting to string together the most common threads brought up by my guests in the first eight interviews on the podcast. Reflecting on these conversations, I'm suggesting some grand challenges for the field in the years to come.
Welcome to the face space invaders podcast, where this time I'm not actually interviewing anyone. This is just a wrap up episode to talk about the last eight interviews I've had with researchers on this podcast. So each of my interviewees brought up important topics and pressing issues that touch our community. And I felt it's a good moment to reflect on the common threads that emerged from those conversations. First of all, I'm really, really grateful to all the listeners who take their time to tune in to listen to the conversations. And I'm even more grateful to my guests for their generosity with their time. I know everyone in the field is extremely busy at the moment. And some of the people I've talked to were really, really patient over multiple attempts due to technical reasons that came up. And, um, I've heard plenty of words of encouragement again from both listeners and guests. So I do have it in me to keep this thing going on for a good while ahead. To the actual questions I think the topic that came up the most in the conversations was the convergence of simulations and experiments. And this really marks an important moment in the development of computational biophysics and biology in that, as we've said a few times, we are for the first time in a moment where we can have predictive power, where we can actually go combine all the biological data that is out there and and built an understanding that is complex enough to actually address a biological problem. Of course, there are also ways around this. As we talked with Alexi, for example, you can go, you know, you can choose to work on a problem that is more industrial and biological so that you can reduce your system's complexity to essentially a molecule and perhaps an inorganic matrix. But if you want to address biological problems, you You have to think about all the networks, all the downstream effects that small modifications on a structural level can have. And this is where the first I'm actually enabled by the amount of data, the amount of literature, the complexity of the tools. So this is an amazing moment to take advantage of this possibility and really, really take your scientific projects to another level in terms of what we can do. There is, of course, a multi scale aspect of the simulations, which, uh, which is based upon the idea that, you know, local modifications have implications for large scale assemblies in biology. That is a very trivial thing to say, but I think it became really, really, um, obvious with the discovery of those huge gigantic assemblies in the cell. And I think it is becoming more and more obvious from the literature that life operates with those mega machines, with systems that are easily made of millions and millions of atoms. And, uh, those operations can rely on tiny details, like macromolecular recognition can sometimes rely on a single amino acid or a few amino acids. It's difference between two isoforms and, uh, I can't help but think of chromatin organization where details as small as single modifications on histone tails can have enormous downstream consequences for how a part of the genome will be expressed or organized or modified. And there are of course many other examples of this, but yeah, this connection of tiny modifications and large scale effects is perhaps one of the biggest challenges that we have to resolve in the field. The other challenge is the treatment of ensembles So the notion that whatever experimental data we have, it's usually an average of many, many compatible structures and different elements of the ensemble can have different functional roles, which makes it particularly important to model things. in the statistical context, as well as really thinking about how the experiment is done in its mechanical detail. Um, this is also a discussion we had with Pilar about sometimes making corrections to our calculation based on, imagining the physics of the setup and so this is where we are in the history of biophysics. Thinking about the conversation with Modesto, we are at this threshold moment where we have these new toys where we have these new capabilities and we really have to reinvent ourselves. It's actually something I myself feel about. how in recent years I was sort of clinging to the old techniques, the old ways of doing things. I was thinking, Oh no, you know, my, my knowledge from the last five or 10 years is becoming obsolete and you know, the way I thought about biophysics is becoming a thing of the past, perhaps, but actually thinking about the capacity to approach more complex problems it's a great way out of this mental cool, this, uh, so we're not really losing things. And this, all these conversations made me realize this with much more depth than I could, I could think myself, we're actually gaining an enormous power and it's just up to us to learn the new ways of incorporating those techniques into our everyday work. And so maybe we need to stop thinking about ourselves as just Simulation people are just structural people or just by informaticians and, uh, start merging the other approaches again, whether that's by informatics or big data or structural AI and so on, it doesn't have to be, you know, you yourself learning every possible technique, but it can be you teaming up with a few colleagues who have this knowledge, maybe learn some of this knowledge directly, maybe, have someone who can always help you with a new branch that emerged recently because someone will always know way more about it than you do. So that we can in the end squeeze out as much information, from these models as possible without confining ourselves to any particular corner of the field. A related question is that of credibility, because as it was said many times, we, you know, we talk to experimenters and they talk to us in the language that we don't always mutually understand, but in the end we need to care that the ideas that we convey to our experimental friends are credible that um that our field can be viewed as contributing to solving a biological problem and not just a nice addition to have you know nice visuals that we actually understand what they are doing we understand What the biological assembly is doing, how it works. I think this requires concerted effort from the whole community to address all the questions of statistics, of model quality, of, you know, having an awareness of the errors we commit with different approximations and the different things that we leave behind or leave aside in approaching every new project, every new biological system. The second topic that came up a lot was that of code maintenance and code interoperability. And I think this is something that has been a problem for a long time. If you ever tried to recover a piece of code that was written 10 years ago, there's a very good chance it will be impossible to run now. And, um, with the advent of GitHub, I think this made things much better in that we have very easy access to publish code. We can publish Jupyter notebooks with examples, but what remains a problem is interoperability within each personal toolbox, I guess. So, one observation I have from my own research is that if everything becomes a separate code environment, then these things become unmanageable again. So, managing conda is a very common complaint. I don't know if we have a better solution, because I myself have probably a dozen conda environments on my cluster that each of them is taking an enormous amount of space They have a crazy amount of files and this dependency conflicts is something I think it's not going to be sustainable. I don't know if anyone is working on this or if anyone will come up with a neat solution soon, but that is something that would be very, very welcome I don't know. Maybe it's a common dependency framework for a field such as by informatics or structural deep learning. So that, you know, you have a common base and only a few libraries would have to be reinstalled every time. But, uh, I would say that's still very much an open question. Another obvious thing is that of fixing bugs and keeping code up to date. So there are many codes that were published at some point in time and then there's nobody to take care of them once the original developer is gone. And this is something that again doesn't have a very clear solution, except for maybe this kind of code consolidation where codes start to be written in the context of long term collaborations, perhaps large groups that have more temporal stability so they won't disappear overnight, perhaps building codes into existing libraries. And of course, this has the same problem of maintaining pieces of libraries while people move on, but this might alleviate this a little bit. if the codes are well documented and publicly hosted. There will be an incentive for future contributors to take them over. I don't know. Of course, we, we need to also think about credit attribution as Max pointed out so that this kind of work doesn't go unrewarded. or unnoticed by the community. And, uh, here, I believe preprints can be an amazing tool to track the popularity and adoption of new versions and modifications of previously published code. Code interoperability in turn, uh, as we talked about, for example, in the episode with Rosen there's a great amount of space for people who want to provide the sort of glue between different pieces of software, because people start to work these days in terms of We're switching from single simulations to dozens of simulations, hundreds of simulations that have downstream analysis workflows. Like if you want to analyze, I don't know, cryptic pockets, allostery, electrostatic surfaces, flexibility, levels of disorder, and so on. All of them can become separate packages of code, but then the question is, are the formats compatible? The ways that things are packaged, are the other outputs easily processed? As inputs to the next thing, can you swap one software for another and keep the same level of simplicity this kind of questions and, um, really thanks to some new large scale collaborations, this seems to be an increasingly fruitful effort in the community. But okay, what about the questions of collaboration and social life? That also came up a lot. And, uh, here we have two topics that are kind of related by I think separate. So collaborations are obviously amazing. We should make sure that we work with other scientists as much as possible. There is an enormous value gained from having a different pair of eyes look at your problem and a different pair of hands work with a different set of tools on the same problem. But perhaps thinking deeply about these questions, collaborations don't just happen. You know, we, we can all put some effort into making them better by practicing our social skills, by showing up to team meetings with results, ideas, and questions by, um, you know, having a clear structural responsibility for the project. So what happens if a project stalls, who should restart it? And, um, yeah, there's wisdom in learning the corporate tools of management and accountability. I think that's something that Vojtech brought up, not necessarily to torture ourselves. And each other, uh, with, you know, administrative stuff, but rather to make sure our work doesn't go to waste and we can have some efficiency in the way we approach science. And then, uh, it's also not all scientific. Let's just remember that. I think there's a point to be made about keeping us sane by making sure that we have a social life outside of work with like minded people and so finding ways to connect through music, art, sport, and culture. Cinema, uh, without even mentioning projects and results and hypotheses, you know, I think every community of scientists should think about these outlets for genuine human interaction. We do that here in Barcelona. In the conversation with Paul, I mentioned that we have a rock band, a football and basketball league, a baking club, a running club, and you can make friends across topics, across career stages, across cultures. And honestly, it's an amazing experience. So then the one topic that is, I believe, on everyone's minds today is that of industry and academia. And it's understandable for financial reasons, but also the pressure that comes from extremely competitive environments. Um, it's good to hear that there are many, many stories of people going back and forth. So people who left academia, for example, after the PhD, they went to do a PhD. an industrial postdoc and then after a postdoc in the industry, they were offered either another postdoc position or directly an academia job. And maybe they kept some engagement with the industry, you know, so it's really, really becoming porous and going both ways. There are some dangers, I believe, in kind of blurring the line between academia and industry, because Obviously, they have different incentive structures, and we know there is a credibility crisis, which is probably exacerbated when purely academic results, start to have a financial incentive to them. But at the same time, there is a great deal of knowledge, as we also discussed in the industry about how to structure projects, how to make sure that they work. things don't go to waste, how to attract people with great knowledge and put them in a place where they can really, really contribute to solve problem in a collaborative way. Because I think academia can often feel lonely, you know, like very often we get a project and we are left alone with this project, maybe being encouraged to, Oh yeah, talk to people around you, but you know, you're still managing a big project without structural support. And the thing we can learn from the industry is how to make teams, how to assign credits to teams, how to make sure that it's not just, um, one person that is charged with all this responsibility for, for maybe, uh, what would be a multimillion project elsewhere But also we kind of share the joys of finding out, we share the credit for. making a breakthrough all this is, I think, an open question and yeah, we can't really ignore the financial reality out there, right? So if there is any way to combine a part time industrial job and part time academic job, I think this will get many, many people closer to, realizing the dream of becoming a professor of being an academic, but also sustaining the families and not having to give up other dreams in their lives for academia. I think it's important we think about that of course, this will be very dependent on the country. Uh, we are often talking about things from the perspective of. Western Europe or the US where perhaps the competition for academic jobs is much higher. I am myself really, really happy to eventually go back to an academic position in Poland and, uh I think there are many reasons to do that because you can have a huge impact in a place that is, you know, full of venture capital, full of people buzzing with ideas. But there's also value in coming to places where this field is not as saturated to bring your expertise and looking at local problems and looking at the environment, the community there and, um, hopefully making a change, you know, help people who are fresh out of university, find an exciting project or even found. An exciting startup work with local hospitals with local databases in implementing things that would never become a thing if, you know, if maybe you weren't there. So I think we need this diversity of people, both going to places that are really hot spots for new ideas and people carrying their insight and carrying their knowledge with them to places. where such a knowledge is really, really needed for developing societies. And so finally we had quite a number impressive stories about how people become world renowned experts in their respective fields. I can bring up the conversation with Giulia about, you know, how this kind of serendipitous discovery of a new thing makes you think about a new project. And, uh, So you write a proposal on this project and then people might say, Oh no, you know, this is not really doable right now, but this question of how people become experts in their fields is really, really a great question because we can see with all examples that many of those success stories. Come from this unreasonable optimism about the project that is, yes, is not doable at this time, but the insight there is that in five years or so it will be doable. This is a sort of point of this whole podcast series that I want to get people inspired about where we're going to be in five or 10 years. And, um, I think having this perspective of what are the opening pathways of discovery it's a great insight that can guide your future career. So I don't know if this is controversial, but I actually love to listen to success stories and the stories of how people got to be where they are. Of course, there is a survivor bias. So don't think that every person who just jumps on bold projects will suddenly become a world renowned scientist. It's not as simple, although that would be amazing But one of the roads to groundbreaking discoveries is exactly that of embracing the next frontier of finding out that, Oh, you know, in five years, we will have either software or hardware, or even the experimental understanding to tackle this next big problem in biology. And if at this moment you can start working towards this, then maybe in five years where this actually happens, you will be the only person in the world who knows how to solve these problems or how to treat those problems. Or maybe you will be the one who made the first tool because there are no tools address this before. So this question of how to find great projects to work on is, is really something I'm. thinking about today. And I'm really grateful to my guests who shared their stories for bringing up this insights. Okay. That's it for this short solo episode. I hope this commentary was somehow helpful for you to bring this information together. I'm definitely looking forward to the next set of interviews. We've got great people invited, and I hope this will result in amazing discussions. Um, as always, one question to the listeners. Please let me know what you find inspiring, what you find nice about the podcast. Is there anything that you would rather skip? If you have people you want to recommend for future conversations, I'm also open to hearing these suggestions. I kind of live in an environment that was shaped by my past interactions on conferences, different meetings. So I obviously know who I want to interview, but this doesn't mean these are the only people who have something interesting to say. So yes, these insights are always very welcome. Well, I hope to see you in the next episodes of face space invaders. And, uh, well, until then, peace. Thank you for listening. See you in the next episode of Face Space Invaders.