Episode 24½ - My thoughts on season 3

Transcript

Milosz: 0:00

Okay, so it's time to wrap up season three of Phase Space Invaders. We've again had eight great conversations in the season and, just to summarize briefly, here are my personal takeaways from each of them. So first we had Lynn Kamerlin, who told us to be brave and venture into interdisciplinary projects and to think of science, maybe less in terms of your field specific toolkit, but more in terms of being a natural philosopher So finding out what fascinates you and what else you can learn to contribute to an interesting discovery, especially when you collaborate with other people and put some time and thought into learning how they think and what they care about. Then we had Erik Lindahl who talked about focusing on solving actual problems. So not just doing some research because you think it will make it into a journal, but actually having an existing problem in the back of your mind in a way that ties you, ties your work to a particular research process going on in the world. And also highlighting this idea that behind simplicity, there's a lot of hard work and just to bring about a simple theory, Very often means spending hours and hours thinking and reflecting about how we can simplify that, how we can find alternatives, how we can, you know, refine our models. Next yet was Alex MacKerell, whose I believe main point was that simple physics can get us a long way. And I think it's a very reasonable voice in the era of AI where we're using increasingly complicated and hard to interpret models. to explain physical behavior of molecules. Of course, there was great success with AlphaFold, for example, which indeed relies on quite complex data processing structures or information flows. But in the end, it might be that we just need more hard, focused work on squeezing the most of our simplest and most reliable models. Because that's how we can keep the human interpretability and understanding that we might otherwise lose. And, um, I think we all observe the complications that come out of over reliance on black box models. This is not to say they will go away, but we also need to think in alternatives. We then had Rommie Amaro, who shared this idea about, um, an open ended exploration of complexity. So starting with a broad question of what if you put together, you know, several components that have never been put together in a model. So what comes out of this kind of open ended observation of complex systems feels almost in opposition to the approach by Alex, of like staying interpretable and focus on the physics. But I think it's really not the position in the sense that they're both great ways of approaching a more complicated, more biologically relevant problems. So one is by focusing on the tool sets, making sure that the methods that we have are absolutely precise and well suited to approach our questions. And the other being to actually bring about the realizations of complexity and the levels that are relevant to biology, because very often we we don't really do that. In episode 21, we had Tamar Schlick, whose main point was finding the right approximation for the problem that we can tackle at a given moment. So the movements of new ideas in biology kind of unlock different problems we can study, such as, you know, new experiments that bring new data or higher amounts of computational resources. And this allows us to simulate new processes. In her case, for example, organization, which But then we need to think really deeply and with some mathematical sophistication about what is the right level of approximation or right formulation that allows us on the one hand to use this data to extract more than we're just putting in, but also does not ignore other key processes or features that we might not have insight into or data on. And this balancing act of approximating just to the right extent is very central to computational biology and biophysics. We then had Lucie Delemotte. And we, of course, had the whole conversation about enhanced sampling techniques and how we still don't understand the specific conditions under which different methods will work. But I think the more salient point of the conversation was exactly to not just into thinking about models of publishing or disseminating. How are we approaching things like evaluation, justice in publishing, The idea that you're only judged on the basis of scientific merit. What is scientific merit? Then how do we judge people when we hire them versus how do we evaluate organizations and units? And what are the different relevant granular levels? What is the amount of volunteering that we can rely on versus actual paid work by people like reviewers and editors? I think all those questions came out. I'm very clearly out of that conversation. Then in the conversation with Zoe Cournia, we focus on the huge amount of work ahead in precision cancer therapies, precision medicine. So Zoe shared this idea with us that there is such a huge amount of proteins and potential drug able sites that have now been enabled. By the latest tools and experimental approaches that it opens up really so many possibilities in precision medicine for cancer They might actually keep us employed for many many years trying to find the next therapy that targets a particular mutation her Her point being that we can now focus on small molecules that address specific cancer mutations and many of them are repetitive I to the level where it allows us to follow this hierarchy of frequency. So to approach the most common molecular origins of cancer first, and then go into less and less frequent molecular pathologies, trying to find the therapeutical modalities for different classes of patients. And the last but not least, we had Katarzyna Marcinkiewicz, the senior editor at Nature Communications, who talked about how we could more often consider writing and publishing as part of the scientific process. And, um, she highlighted the fact that there's a lot of science that actually happens after we have our manuscript prepared and sent out for initial review. So we get this human feedback. in the end, many editors are focused on is improving the quality of science, and we should perhaps be more receptive to the well intended, often well intended push for improvement that comes from both editors and reviewers. And so these were the latest conversations. And since it's also the beginning of a new year, I thought about the short history of the podcast itself a little bit after almost one year of publishing episodes. It's actually more than one year by now because the story of the podcast, as I alluded to already in some conversations started around August when I went to EPSA, the European Biophysical Stockholm. I stayed for a few extra days visiting Lucie Delemotte lab. And I talked to my friend, Antoni, about this idea of the podcast, refining the concepts, um, thinking about how to organize it. I reached out to the first guests around October, November, 2023, and it took a really long time to take off because initially the platform I used for recording and editing didn't work well. There were technical issues and I couldn't really afford to have a platform that's not reliable, you know, inviting and talking to people who are extremely busy and sometimes, you know. They can only schedule one hour for me in three months time, say. So I had to switch the platform to work on the concepts and introductions and, you know, record the jingles. I still remember. the times in January where I was sitting here and thinking about, you know, how I'm going to structure this whole thing and, um, isn't even going to, to work out. And, uh, so yeah, one year from then, I've got 24 episodes, 24 interviews. Each of them was great, but hell, were they stressful. Some of them obviously are with friends who I have known for a long time, but many times I'm talking to people for the first time and I'm still amazed by how many guests are willing to dedicate their precious time to an initiative that has started so small. Um, I'm really, really grateful for that. I always highlight this and I've also had great feedback from the community. So Even my guests sometimes say, you know, I was thinking of doing something like this, but I never had the time. So thanks that you're doing this, which is fair. I still believe I'm, you know, among the least qualified people to do it, but at least I can credit myself with taking up the commitment. That's something I'm also getting great support from the listeners. So I received a few messages from, you know, platforms like Twitter or Buzzsprout, even through direct emails from people saying, Oh, how much they value this podcast. And I honestly couldn't be more surprised and humbled by this feedback. It is incredible. I'm really thankful to every person who reaches out to me to tell it's worth my time. It's worth the effort. It really gets me going. I also want to weaponize this support a little bit. Jokingly, of course, to tell you that, you know, referring to the conversations we've had both on air and behind the scenes, that there are just so many impactful projects that ambitious people could embark on. Of course, I understand that everyone in the community is extremely busy. It's always been like this. And that's a major concern that any new initiative has to be put on top of the existing things that people are doing usually for free, but it's not I anyway want this message to be heard by people who maybe look for something impactful or ambitious and don't know where to start, don't know where to direct their efforts. So one thing we discussed with Alex MacKerell was the question of reference data for the development of molecular models, including force fields. And there are really, really many experiments that were done in the sixties or seventies, That were simple physical chemistry experiments, osmotic pressures, free energies of association for small molecules that are biologically relevant. Many times those data sets are missing or conflicting or partial, or maybe we don't trust the methods that were available at the time. And it's really hard to find data that is more modern and reliable. So if you can identify a simple experiment that can provide additional insights into how to train a force field, probably in collaboration with someone who works on that, that can be literally a major improvement in how we simulate, for example, nucleic acids. We are working ourselves on some things along these lines. Um, again, a small contribution, not to spoil too many ideas, but sometimes it is the cheapest thing that you can do. if you can still get funding for that, that gets you probably the longest way. It's also understood that many of those data sets are somehow private held by corporations or pharma companies, but the impact of having those data sets publicly available cannot be underestimated. With Lucie then we had this other conversation about lack of benchmarks for the hand sampling methods. And that's another thing that seems like an obvious thing to do. Of course, different people work on different systems and they will have different ideas about what is a good benchmark and what is not. But perhaps getting together a number of people and working on something that is standardized, works across many scales from the simplest to the most complex, and has very good or known a priori reference values, like in the case of symmetric systems, or slow kinetics that can be unambiguously measured. And then having a collection of some systems with inputs and parameters and showing that for a few enhanced sampling methods, these are the ones that work well. These are the ones that don't work at all. And, um, you know, here's how to test your own method. This could be a really valuable resource for the community. again, I'm not saying, you know, do it on your own. Um, I'm saying we need people who will bring other people together to make it happen and will be able to dedicate some of their time to, to, to make it happen. And another thing that I wish we had as a field is perhaps another podcast or maybe a YouTube channel that would do, I don't know, a weekly journal club about the most interesting computational biology papers that come out, or even a review of historically important papers in the last three, four decades. Um, I know this is something that each group does on its own, and yet it seems to be such a low hanging fruit to make it public. Although I know it sounds easy, but it actually takes work and dedication. Um, but some fields have those people who do that. I think our field, as far as I know, doesn't. If you're thinking about that, you know, get in touch or just start doing it. and then there are possible projects regarding, you know, higher level ideas. So Of course, publishing models are a big thing. It's hard to tackle this as a single person, but yeah, we've mentioned initiatives like science collab or biophysics collab. This is one place you can go where the idea behind this is that these are people who provide reviews for preprints. Um, of course there are way fewer volunteer reviewers than their preprints. So this is not going to completely solve publishing, but Any new and fresh idea for how publishing can improve can be valuable. In the conversation with Katarzyna, I mentioned this idea of a free market of papers or some sort of community portal where people would be able to provide feedback and suggestions and scoring and you know, signal interest. I think even coupling that to language models to give hints about I know strengths and weaknesses would be an interesting experiment. This is risky, but, um, it can still work to some extent. People could in principle design such projects almost from scratch without funding from major publishers and still have an impact on how science will evolve. And so here I want to reiterate the bottom line that. Many people highlighted or I highlighted in the conversations that in the end, the structure of publishing the structure of science dissemination will depend on us on how we agree that science should be reviewed, published and communicated. It's not something that is being forced on us by by the government or by the publishers, maybe to a small extent in the sense that, oh yeah, it's good if it has a number on it. Um, it's good if it provides income for, for the publisher. But the bottom line is that if we come up with a great idea, everyone will be happy to adopt it. And when we are talking about platforms, there's also this question that this podcast tries to address in part, which is focusing on higher level concepts in science. So topics that are, let's say, meta scientific, some of them can be discussing our paradigms. Some of them can be discussing the possibilities that are brought about by new methods or new hardware and software developments. obviously, things like publishing or hiring or the academic market. I would say that we still I still don't have a unified platform for such discussions. I have in mind something like Kialo, for example, K I A L O, which is a platform for generic debates in philosophy. We had several guests who routinely write perspectives and reviews to integrate the knowledge and paradigms within each particular subfield. And these are published in many, many different journals. So we could. Um, I think it would probably use another level of integration where these perspectives, you know, would be put into context more explicitly. That's a maybe far fetched idea, but I think people would love if something like that existed. it is great though, that BlueSky is again emerging as one such platform. So if anyone is not there, I am now on Twitter and BlueSky and LinkedIn, but among them, I would say that, Blue sky looks the most promising for the scientific community. So I encourage everyone to maybe not move there fully, but at least be there and take a look at what's happening and see if you like it. Um, of course, eventually it would be amazing if we could all agree on one place where the community would sit and thrive and, um, have those conversations and kind of organize organically. By the moment, maybe we need a bit of unity in terms of deciding what is the right format for such discussions. And what I want to get across is that anyone who has inspiring ideas or great ideas on that can, in principle, make a big impact here because, you know, in the end, Science is a social enterprise. And I think this podcast also tries to highlight this, whether it's evaluation of science and getting inspired or learning or getting personal feedback or, you know, getting this pat on the back that Tamar alluded to, you know, great job, good job. We need the community. Um, Of course, at the same time, and this can be an easy critic of my conversation with, with Katarzyna, we need. to keep the objectivity and universal standards. So that makes the whole thing, the whole enterprise more complicated because obviously there are people within the community who are famous and known and respected members. And history knows cases where these people's words was taken at face value, even when they were wrong. And obviously there are people who maybe have great ideas, but nobody has heard about them at all. Um, so in all those new systems, new concepts, we need to balance those two extremes out between fully relying on community and opinions and personal preferences and being able to Objectively, maybe anonymously evaluate ideas on their merit only. Um, yeah, I'm going on a bit of a tangent here, but I myself recently leaning towards this so called metamodern concept of both and in approaching complex social or societal problems. And I invite everyone to try this lens of, as I said, both and thinking, which means that we both need to acknowledge, so to say, the game or the system or any kind of complexity that emerges naturally from people following their incentives and try to genuinely improve things by designing better incentive structures. I recently heard the great phrase, you know, avoid game denial and game acceptance So game denial would be, for example, assuming that we're all just motivated by the best science, we should all be well funded and publish the best versions of our research, and it will all work out naturally. Uh, because in reality, we will always be evaluated and compared and somehow organized into rankings and some science will be more impactful than others. And we are all employed by the society or by many societies that allocate limited resources to our work that they could spend also somewhere else. And then on the other hand, we cannot simply agree that, okay, you know, we have those big journals, we have impact factors, we have here in this is, and that's the ultimate truth about your research that we cannot be just, you know, complicit or satisfied with the simplistic tools that we have. So we need to use our most inspired or transcendent values such as objectivity, truth, justice, humanity, to make sure that whatever we come up with addresses those questions in publishing, in hiring, in grants, in communication. And instead we might get really bogged into those false dichotomies that exist on social media in many debates, rather than trying to find a solution that can address the underlying concerns of both opposing sides. And very often this can be done with a bit more refined thinking. So sometimes we need to abandon our simplified concepts. And maybe as we talked with Lucie, We'll have a different solution for hiring and a different solution for evaluating and a different solution for grant applications, right? Maybe they just need three different structures and there's no one size fits all approach that can evaluate the researchers. We've heard in several conversations. So more bespoke solutions, more fine grained thinking can perhaps get us a long way and help us avoid many of those exactly false dichotomies that we often fall into. Okay, two personal advertisements towards the end if you've gotten this far into the podcast. Um, one thing is that in the long break between the episode with Alex MacKerell and, uh, we had Rommie Amaro in the meantime, but Alex and Tamar Schlick later in November, This was the time I was visiting the VMD development team in Urbana Champaign working on a plugin for VMD for my movie making tool called Molywood with a single L. So VMD is now releasing its version 2. 0 and Molywood will be available as a plugin for making movies directly inside VMD. Here I have to extend the huge thanks to Barry Izralevitz, Emad Tajkhorshid. And Diego Gomes, who has also been an incredible supporter of the podcast. I hope this plugin is helpful to those of you who maybe missed such a feature in VMD. I will be publishing some snippets. And examples very soon. So, you know, stay tuned. If you, if any of you wants to test it and give me feedback, please feel free to reach out and, you know, let me know if there are any features missing, if it works for you, or if you find the idea useful. The other thing is the financial disclosure. So I'm a working academic who is paid for doing science, and that's amazing in itself. And I'm not trying to make an income from this podcast, but I have costs of the subscriptions for the editing and hosting tools, which amount to something like 300, 350 euros per year. This is not a major cost, but I thought I will start a buy me a coffee page. And if any of you wants to chip in, you know, one euro or dollar per month, I'm extremely happy with that. If I have any excess money from that, I will funnel it into a bit of advertising. I see from my hosting provider that actually 90 percent of the podcast audience lives in Europe and the US, which is great, but I would also want to see if there is interest in say Latin America, Asia and Africa, especially among people who might not realize that there is a global community around our field. And I also acknowledge that our community At least my selection of speakers is obviously Western biased, this might be an obvious reason why we're mostly reaching Western audiences, and this is something I want to change eventually. But for now, if I have some extra resources, I would probably try to reach the people in those regions where perhaps there's a huge number of people who want to learn about the field, but they are not reached by the algorithms or the modes of dissemination that I'm using today. That would be a fair use of any excess money that I would make from your contributions. And, uh, okay. With that, I want to thank everyone for the support I've had throughout this year. It's really been more than I could imagine at the beginning. I have been to some conferences in person. And people told me that, you know, they heard about the podcast, that their listeners, they share their favorite episodes. Someone actually recognized me by my voice, which I never thought would ever happen to me. So again, I acknowledge it has been an absolute surprise. I've learned a lot, not only about podcasting and those technical tools, but also about science and possibilities. And that's shifted my own approach to computational biophysics so much. And I hope you're learning together with me. I think our guests bring incredible points to the discussion, and it's amazing to have these personal insights into how science develops. And I will be coming back with more episodes very soon. I will take maybe a few days or weeks off and I'll be coming back with season four for another eight conversations. I've already got some of them scheduled lined up and I know these are guests that you will absolutely love. So I hope you're looking forward to it. So I wish everyone all the best in 2025. I know as of now, the world is a messy place, but I can only wish all of you this stay safe and healthy. Get your fulfillment by aiming for some great scientific achievements and find plenty of inspiration to keep you going throughout this year. I'll talk to you soon.

26:09

Thank you for listening. See you in the next episode of Face Space Invaders.