Episode 11 - Justin Lemkul: Providing technical help online, sharing expertise, and polarization in nucleic acids

In the eleventh episode, Justin Lemkul and I talk about the motivations and challenges behind providing technical help on online forums and mailing lists. Justin shares his story of becoming a prolific technical advisor for the Gromacs community, which evolves into a discussion about automatability, the role of expertise in running and documenting simulation-based projects, and the incentives for people who contribute their time to helping the community. We then move on to discuss polarizable force fields for nucleic acids, including the problems they might help solve and the practical ways in which we might arrive at these solutions. We conclude by pointing out the need for more nuanced undergraduate curricula that reflect the current open questions in nucleic acids research to attract students who will make the future discoveries.

Transcript

Milosz: 0:00

Welcome to the phase space invaders podcast, where we explore the future of computational biology and biophysics by interviewing researchers working on exciting transformative ideas. Today I'm talking to Justin Lemkul, an assistant professor at Virginia Tech, working on the development and application of polarizable force fields for nucleic acids and amyloidogenic proteins. On top of his very interesting research agenda, including non canonical structures, Many of you will know him from his activity online, where he contributed to tutorials, manuals, and answered an innumerable amount of questions on fora and mailing lists. So we talked with Justin about how he started his journey as a community problem solver, something he traces back to his own struggles with the first simulation projects. We discussed the tough balance of ease of use versus expertise, questioning if there can ever be a true automated system. for running molecular simulations then Justin moves on to make a case for better treatment of polarizable electrostatics for certain molecular systems such as the highly charged and dynamic single stranded nucleic acids we end up by wondering what it will take for molecular simulations to make real contributions to the boom in RNA research, and how we can make sure that the next generation of researchers is on board with us there. I know you can't wait, so let's go okay. Justin Lemkul, welcome to the podcast.

Justin Lemkul: 1:46

Thanks for having me. Appreciate the invitation.

Milosz: 1:49

So for many people, you are like the go to problem solver of everything computational. You know, that's a, that's a well earned credential. just on the two websites I myself follow, which are ResearchGate and the Gromacs forum. I've seen, you've posted advice over 4, 000 times, which I find mind boggling. And you also created a set of Gromacs tutorials that help so many people do their baby steps in simulations. And I'm curious, first, what prompted you to commit so much effort to sharing technical knowledge online? And, um, if after all these years, do you have any takeaways on what works and what doesn't when it comes to, you know, online tutoring?

Justin Lemkul: 2:32

Yeah, it's, it's been an interesting path through that. I mean, when I started using Gromacs, there were very limited resources for, for learning it, you know, and that was, I had used Amber software before that to sort of get my introduction into MD simulations. And then I moved to Gromacs for a different project. And, know, a lot of it was with Amber was from my research advisor. He had used that software. He taught me how to use it. And then when it came to Gromacs, he had never used it, but he said, Hey, this might be faster, might have the features we need or whatever. So let's try to figure it out. and I recall when we started. There was myself and a couple of others in the lab. There was really like one tutorial online for using Gromacs on a protein. And it was really just a glorified list of commands, which I appreciated because that was the workflow, but there was very little, um, you know, instruction that went along with it. Like, what am I really doing in each of these steps? And we had to kind of piece it together. And so I started using the Gromacs forums myself as, as someone asking for help, because unfortunately the system I was working with, the first thing I ever did in Gromacs was a membrane embedded peptide And that was very challenging for someone who was still fairly new in MD. And so I had a lot of headaches and, and a lot of very kind people helped me out and, and I got my project going. I, I just sort of realized there's this gap in our field of, you know, the, the essentiality of having your advisor know how to do something to teach you how to do it. And I thought that was going to sort of hold us back. And I just started documenting everything I was doing really thoroughly, which, you know, I think that's just good scientific practice, but you know, then trying to say, well, these are my notes and they make sense to me, but could I actually use them to teach somebody else? And. Then I, I started getting daring and answering people's questions. And that I tell you that the first time I responded to someone on the Gromacs, then, you know, it was a mailing list. It took me a while to hit send on that email. Cause I'm going, this is a real easy question. Like you actually do know the answer to this, right? Like it, it kind of scared me. And, uh, if memory serves, that was around 2008 or so when I started actually helping other people. Honestly, it just, it feels great to be able to do that and answer people's questions. And so became rather prolific at it, I suppose. thousands and thousands of emails and research gate posts and now forum posts for Gromacs and things like that. And it just, you know, was sort of, there was a void. And then I started documenting it online. I said, I keep seeing the same questions over and over again. and this is where some of the documentation, got improved. Actually, Eric Lindahl reached out to me and said, Hey, would you want to help out doing some work in the manual for Gromacs? Because, you know, you keep addressing these things. And I said, sure. You know, that sounds great. I was really honored by that. and I did a lot of writing in the, what at the time was, you know, the, the PDF manual, and then some help on the wiki page, which is now morphed into what the current version of things, uh, happened to be. So, know, it's, been a lot of, Interesting challenges in scientific communication, right? You can tell someone, okay, do this, get this, but what is their experience level going in? The absolute novice versus someone who is somewhat familiar or someone coming from a different field like quantum chemistry or something. They know a lot of things, but maybe they're, they're viewing things differently and just trying to communicate things as as understandably as possible. I think is really the core challenge, to mention things like language barriers and technical capabilities and all of those things that you run into, Which is a whole story unto itself. But a lot of lessons learned over the last, I'd say, 15 years or so.

Milosz: 6:08

Right, right. I'm just thinking this was very much on my mind when I was starting this podcast. Because the question of the people who come into the field without. direct supervision, right? Because I had also a great supervisor who taught me all the basics. And I, I always had someone to turn to when I was in doubt about some, feature of the code. but then there's plenty of people who are not this lucky, right? And the one thing that bugs me a bit is that all sorts of questions on the forums repeat, and this are sometimes, you know, basic questions that were addressed. thousands of times. The question is because we're probably trying to build more robust workflows now, things that are more automated, uh, do you think there's a way out of this or you know, is it always going to be working with people who have the first error in their first simulation and they turn to ask the question, what can be done?

Justin Lemkul: 7:06

Yeah, I think, I think it's a blessing and a curse of having really strong online communities because you say, Oh, well, there's a forum. I'll just go ask when in reality, I would say 90 percent of the problems that people encounter are answerable with a simple Google search. it actually blows the minds of some new students. And I see this with new students that I work with. Sometimes they go, wait, I could just like Google it. And there's an answer. And I'm like, more than likely, um, the really corner cases. Those are the fun ones to work on. Like, wow, why did that happen? Like, is it a bug in the code? Is it a false assumption? We're making something. Those are the ones that are fun to kind of dig into. But, you know, why did PDBDGMX tell me that residue UNK cannot be found? I mean, come on. Um, it's,

Milosz: 7:48

Yeah.

Justin Lemkul: 7:49

of funny to see that one continually come up. But yeah, you're right. People are looking to do more complicated things, and they're increasingly being asked to just go and do them independently. You know, my word of caution there, and the thing that I've struggled with, because I empathize very, very strongly with these people, but you can't always help everything, is I get emails all the time. About my supervisor doesn't know how to do simulations, but I've been told I have to do this. there's no one here at my university that can help me. I'm asking them the question, well, why are you doing it? And I don't mean that to sound rude. I know some people take my messages sometimes as being very snarky. I have a very sarcastic sense of humor, so bear with me, but I try to be very direct online because I know there's, it's a lesson learned from language barriers, right? I got myself in trouble trying to be humorous too many times and people were like, what is this guy talking about? Um, it's just to be very direct is to say, if there's no one there that can teach you this. You know, I got a PhD in this field. It took me five years to really master this stuff. And you think, Okay, I'm going to do this really complicated thing just because I'm being told to. You really have to caution yourself about viewing an MD simulation as a black box. It's not. You have to have a lot of specialist knowledge. I think the activation barrier, if you will, to getting in is fairly low with, you know, rudimentary tutorials, like some of the stuff I've written, I hope makes it easier for people to do the workflow. But then as soon as something is a little bit different, And you haven't actually studied the field to know what you're doing, you go, well, now what not now I'm lost. And that remains a problem. And I think it will remain a problem because I see it, you know, a lot in talking with colleagues and, you know, grant reviews and paper reviews is like, well, why don't just add a simulation onto this to make things, you know, Look better or make more sense or whatever. And I'm like, that's not how you like, we're not toys to play with. This is a real dedicated field of science. Right. And I'm sure I'm preaching to the choir, to you and the audience here, but you know, that's, that's the, the, the conception I think some people have is, Oh, well just use your magic computer box to give me a cool answer. And it's like, that's not how we do this. Like it's real science that you've got to think very critically about and so I think we're going to continue to face this challenge and I honestly don't have an easy answer to it. It's just. Continuing to see where the field is moving and what resources can we develop for people And how much can we communicate to those PIs who tell their students? Oh, just go and do this It's simple and it's like it really really isn't And please understand that it's really not easy to run a high quality Simulation. It's it's you're doing real science,

Milosz: 10:21

Right. I hope you don't preach only to the choir though, because I hope part of the audience is people who, you know, who are just beginners in the field. I mean, that's at least a desired target. Yeah. I very often see that the main answer or the most reasonable answer to many questions would be, you know, like, tell me your research project because I don't understand what you're trying to do. Right. So. People often ask, is it okay to do this, right? it might be okay in like 20 circumstances. It might not be okay in 50 other circumstances And unless I'm your PI, I cannot tell you which one this is

Justin Lemkul: 10:58

right? And that's

Milosz: 10:59

So.

Justin Lemkul: 10:59

first responses. Well, you guys, what's in your system? Like what, you know, where's your quarter file? Is it a modeled structure? Is it an experiment? Like, are you trying to modify something? You've got to give us context in sort of asking that question in a really useful way. Because, yeah, like you said, if I tell you, Oh, yeah, that'll work if you do this. And that's the 1 percent case where it doesn't. And then I look like the fool trying to give you advice that won't work.

Milosz: 11:22

Yeah, I very much sympathize with this worry because we've had also discussions here, you know, about, I don't know, automated workflows missing, disulfide bridges, right? So you put something in a workflow and then you realize, Oh, your protein doesn't have a disulfide anymore. And how do you detect this? well, unless you really pay attention to what's happening.

Justin Lemkul: 11:42

Sure. Or, uh, you know, a protonation state or something like that. Easy to miss. And it's very hard to get those things universally. Yeah.

Milosz: 11:50

Yeah, I wonder if we should have some sort of universal checkers that, you know, have all those possible questions put into an AI system or whatever. And then telling, you know, your system might be missing this. Think about that. Here's some resources to read about. Uh,

Justin Lemkul: 12:09

Yeah. But I mean, you think

Milosz: 12:10

yeah.

Justin Lemkul: 12:11

buried residues that might have a different PKA like that. That's one that haunts, you know, all biochemists like some oddball residue where you just go, Oh, well, that's this and I'll, you know, titrate it a certain way. And no, it's essential for the function of the enzyme or something to be different. And don't know how you automate that necessarily, but it's an important thing to think

Milosz: 12:31

Totally. I remember talking to Alex MacKerell and he told me that his solution for those things is just to have checklists for everything. So,

Justin Lemkul: 12:39

Oh, well, and I was a postdoc with Alex and I know

Milosz: 12:42

yeah.

Justin Lemkul: 12:43

his famous phrase of check your inputs, where it's like, do you have, do you have a real strong sense of what you're doing and can you map back to every decision you made along the way to justify what you did? And yeah, if we can teach a computer to do that, that might be helpful, at least to augment our own understanding of the problem. Yeah,

Milosz: 13:03

Right, right. Maybe that's one of the ways to go for the tutorial making and automating communities.

Justin Lemkul: 13:11

could

Milosz: 13:11

Out of curiosity, did you have any projects that came out of, you know, people asking random questions? questions online? Do you sometimes get involved to the level of, you know, actually getting into the science of it

Justin Lemkul: 13:24

not often, you know, in terms of collaborative work or contributing to stuff, you know, I, I try to, my problem is I say yes to too much. That sounds interesting. I think a lot of folks will empathize with that. So I, I try to get my, my satisfaction from just, okay, great. You've, you've now got a direction for your project. I've, I've been helpful in some way. some folks do reach out for collaborations just because they recognize my name or they know that I work in a similar space because they've seen things online and, you know, occasionally some things have come up from that. Um, you know, there's a group that I work in Germany. They reached out to me, I guess, about two years ago now to do some steered MD type stuff. And they had seen my tutorials and they knew about me and, had a very, very bright student who had already taken a lot of initiative to do a lot of the work. And they really just wanted to sort of consult with me and see you know, if things they were doing were making sense. And, and that was, that's been a very interesting collaboration. Cause I've gotten to work on some projects I would not have otherwise done, which are very cool. And we're continuing to do those things. So. You know, I would say every now and again, something comes along. but I honestly get far too many requests for, Hey, could you work on this with us? And it's just, you know, I, I don't have the hands myself and even in my lab to delegate to folks as, as side projects or something, because a lot of them are really involved and, and you just have to tell folks, you know, I can give you some advice here, but I'm afraid that's where my time sort of ends.

Milosz: 14:46

I can imagine. Yeah, I was just thinking that, you know, there's not really that much incentive in the field for something like making tutorials because as far as I know, you haven't published them, right? So they're not even citable. Uh,

Justin Lemkul: 15:00

they are. Um, I

Milosz: 15:02

they are now?

Justin Lemkul: 15:02

opportunity to, to cite, um, the GROMACS tutorials, the, the Living Journal of Computational

Milosz: 15:07

All right.

Justin Lemkul: 15:08

Really cool, uh,

Milosz: 15:10

That's a great journal.

Justin Lemkul: 15:12

great journal. And, and I think they've got wonderful resources. So if people haven't seen that, definitely check them out because these are living articles. And I'm actually was just chatting a couple of months ago with, with Michael Shirts about doing an update to my GROMACS tutorials for them, because. They're a bit dated, you know, they're 2018, so they're, they're a little outdated now, and I'm also working on a slightly different set of tutorials, sort of, based on those, the Journal of Physical Chemistry is going to have an article collection on tutorials, and, and I've been asked to contribute to that, so, there are citation incentives there, you know, I will say that there is a benefit to that, You know, you can collect web server statistics and stuff. And I put those in my grant applications. It's

Milosz: 15:49

All right.

Justin Lemkul: 15:50

where there's an educational spin to it or, or an outgrowth to say, look, the research that I did, a lot of those tutorials came directly from projects I worked on. I mean, the umbrella sampling one was a published paper, you know, and the biphasic system was, was a little piece of another one, you know, free energy things that I did. So if you can demonstrate the impact of that, I've been floored at looking at the Server statistics. You know, I started this, I'm like, Oh, it'd be cool if a few people read this. I mean, there are like hundreds of thousands of page views over the period of a year on those things. It, it blows my mind. So there, there's definitely a benefit to that. It's, it's a nice sort of tangential thing. I didn't do it for that reason, but it's, it's nice to see the impact and be able to say, Hey, this is. This is something that the field needs and continually uses.

Milosz: 16:35

Right. Now, I was about to say that, you know, there's another sort of reward in terms of visibility and maybe, you know, higher rates of, student application for your lab and so on. As you said, people recognize you and want to work on interesting things, but that's also a great point. I mean, the live journal is, it's a great initiative and I love the idea.

Justin Lemkul: 16:55

Yeah,

Milosz: 16:55

hope they,

Justin Lemkul: 16:56

I mean, the entire article exists as a GitHub repository. You know, people can suggest changes, they can make contributions to it via pull requests, you know, that kind of stuff. It's, it's really great. Uh, and I

Milosz: 17:06

right.

Justin Lemkul: 17:06

innovative publication model and, and I think it'll, it'll go a long way to helping the field

Milosz: 17:11

totally. Okay. Uh, let's go back to actual scientific project. So you're working on the development of polarizable force fields for nucleic acids. And that's, you know, really close to my heart based on what I do as well. Can you give us an elevator pitch for why nucleic acids can benefit from being modeled with explicit polarization?

Justin Lemkul: 17:33

Sure. I mean, I, I think that Given the nature of the interactions that you see in nucleic acids, you know, there's so much induction in electron clouds, you know, even in, even in base stacking, you think of that as sort of a generic hydrophobic effect, but there's a lot of induction among those pi systems when you get bases to stack on one another. Uh, and not to mention the sort of huge one, which is like RNA folding and the, you know, coordination of magnesium and maybe some other ions to adopt tertiary folds. These are all strongly electrostatically driven interactions. Um, you can capture much of it pretty well with additive force fields. I'll say that off the bat. You know, I'm not here to say that. You know, there's, there's not good value to additive modeling. There definitely is. But there are certain cases where I think that things like, you know, divalent ions, they're not modeled well by a sphere with a plus two charge. They're just not and if you can model that inductive effect, you get a better sense of the real energetics of the system. and not to mention, you know, even things like hydration properties and the balance of hydration versus. You know, stacking and intramolecular hydrogen bonding and things like that. Water interactions are really well captured by polarizable force fields. And so I would say there's definitely a set of applications and hopefully those little ones we're working in, that really benefit from having an explicitly polarizable model as sort of a new way forward to understanding the biophysical properties of those systems.

Milosz: 18:54

Right, and especially in all those flexible single stranded nucleic acids, right, which adopt really, really crazy conformations once in a while.

Justin Lemkul: 19:03

Yeah,

Milosz: 19:03

I think, looking at your papers, I think quadruplexes, right are one of those structures that can benefit from,

Justin Lemkul: 19:09

are, the quadruplexes are crazy case. I mean, and it was interesting because that was just a complete sort of off the wall idea. We had, you know, tuned the DNA force field to duplexes, you know, fixing up some, some minor problem, well, some problems with our duplexes in the previous version of the Drude force field. And I spent, several years doing that and then refining things for RNA and whatnot. And, Alex stopped by my office one day and said, have you heard a lot about these, you know, these G And I mean, I had heard of them, I was aware of them, but I was so focused on like parameter land and duplex DNA that I hadn't really thought about them in a while. And he said, I wonder if the polarizable model is, is better for those, you know, because they coordinate those monovalent cations in the core. Um, And that has been a challenge with, with additive modeling. Now I know there's some parameter sets that do get that coordination reasonably well, but, we just kind of just tried it out. It was, you know, we picked one and simulated it and I went, wow, it, it's like rock solid, it's really good. Like

Milosz: 20:06

hm,

Justin Lemkul: 20:06

distributions look pretty good. There's a couple that are maybe a little wonky, but the ion retention is perfect. And has become a core component of what my lab does. one of these just random ideas of, Hey, let's try this. has really developed into something very, very cool. And that's definitely an area that I think, you know, the polarizable modeling is benefiting nucleic acids is, you know, those ion interactions and just non canonical structures, right? Cause so much of our DNA and RNA force field work is targeted toward canonical structures, which, I mean, what does that mean for RNA? But definitely the DNA force fields, it's like. wanted to preserve those duplexes, right? and in some sense, those might be too rigid. And so moving forward, you know, looking at a wider variety of systems is, is really cool. And I think it's going to expose some new challenges and new opportunities for refinement and improvement.

Milosz: 20:59

yeah, the non c anonical ones are definitely a frontier now. Do you see other interesting structures that come up that could benefit from polarizability being explicitly treated or is it still mostly quadruplexes?

Justin Lemkul: 21:14

Well, we're, we're getting into the RNA area. you know, I mean, there are, there are RNA quadruplexes, but we're looking at other things in RNA. You know, we've always sort of had an eye toward riboswitches and even ribozymes and things like that, where I think. What's been really well established, or reasonably well established, is, you know, in protein enzymes, electric field effects in active sites, they depend on polarization, right? I mean, that's not even a new idea. 50 years old kind of idea that induction is really important for, you know, for catalysis. so no one's really looked at that a whole lot in terms of RNA, and that's one of the things we're looking at is, okay, function in terms of things that are ribozymes, because they are enzymes and they carry out chemical reactions, but what also about folding? is there a relationship between, molecular dipole moments of the bases or something and electric fields that are being exerted on them by the RNA itself? And we're looking at this in the context of folding of, you know, Tetra loops. Um, that I gave some presentations this past spring at a couple of meetings on this. This is a new idea for us looking at some ribosomes, some synthetic ones, some engineered ones. things and some endogenous ones We don't know a lot about RNA catalysis in that regard. And so I, I think that's a new area for us. That's going to be very cool.

Milosz: 22:31

Yeah, that's exciting because with all the new things coming up from biology, RNA is definitely, you know, looking like the next revolution. All of the sort of

Justin Lemkul: 22:41

definitely an important, you know?

Milosz: 22:43

classes. Yeah.

Justin Lemkul: 22:44

Yeah. Yeah, I mean, I get to tell my students in class all the time. It's like, look, DNA is not this inert duplex that you've been told it is. And RNA is not just this little noodle of messenger information. There is so much going on for both inside the cell. and it's, it's an area really that I think. You know, our curriculum lacks in many ways when we're teaching biochemistry and biophysics and things is you don't have enough time to necessarily dedicate to all of those kinds of concepts, but they're really important because, as you say, they're the emerging ideas in the field. We're gaining a greater understanding of these things and their role in the cell, but they're very hard to interrogate experimentally. And that's where I think simulations, if you have a really good model, simulations can really, really help, you know, augment and drive our understanding of how these things behave.

Milosz: 23:31

Yeah. I was doing a short kind of review recently and I noticed that many kinds of those RNAs have even not. ever been simulated, you know, with any sort of resolution. So it's very much an interesting frontier. And you're very much right also in saying that the curriculums are not ready for the amount of novelty coming out. yeah, I wonder how much we can incorporate. I remember, you know, uh, what was it, 15 years ago when I was going through cell biology, my professor mentioned vaults right? Those mysterious organelles that We're found floating in the cells and they have now people know that they have volt RNAs and nobody knows what they are, but it's kind of this kind of mysteries, right? As you say nobody knows the function. People know something is there, but what does it do?

Justin Lemkul: 24:22

well, and maybe in a few decades time they'll be actually writing that into the textbooks. Maybe we'll have learned a little bit more.

Milosz: 24:28

Right? Hopefully, I mean, would be great. As we discussed before, to convey more of the information. Okay. Thank you. cutting edge uncertainty to the students as we update or create our courses. So, I think it's a conversation we need to have in order to bring more people into the field making discoveries in the next 5 10 years.

Justin Lemkul: 24:49

there's a lot of exciting science to be done there. It's complicated problems, but I think there's a lot of opportunity there. And, and we'll learn a lot about how life works if we get it right.

Milosz: 25:00

Right, I think we need to rethink how, you know, us as simulators can contribute there. Because it feels kind of obvious how it can be done experimentally, but do you have any obvious ideas for like how simulations can you know, enhance or contribute to this revolution?

Justin Lemkul: 25:18

I, I think one of the areas, because as you say, you know, a lot of important, you know, even short RNAs haven't been simulated. And. It's a challenge. It's a bit of a paradox, I suppose, because we can simulate them and then everyone will say, well, how do you know the Because there's no experimental data on a whole swath of these things. You say, well, that's the point. I'm giving you the simulation to predict what it's doing. and hopefully, you know, reviewers and all will find value in that. But I think that's an area we can drive things forward, right? Really good simulations do one of two things. They give you a mechanistic rationalization for why something happens, or they give you a new hypothesis to test. And I hope there's still a lot of room for pure theory studies that say, hey, I simulated these RNAs. Here's what I think their structural, you know, their conformational ensemble looks like. Here's where they differ based on sequence or modified base or whatever. and yeah, it's purely predictive, but In the absence of any experimental data on that, hopefully it motivates people to then say, well, I could test that. I could test that with fret labeling or NMR experiments or or whatever happens to be, um, and there just has to be enough of a sort of groundswell of interest for that to happen. I think this is an area where we have pretty robust force fields. We are lacking in validation data for many of these systems. I mean, there's, there are some that have come up. Of course, people simulate like crazy because there's a lot of really good, you know, NMR data out there, what have you. but we're missing a lot. And I think people need to, to turn their attention to that and start doing experiments. And maybe if we push and say, I've got a bunch of simulations here. I need someone to tell me if I'm right or wrong that's, that's the way to go because. You know, I tell folks all the time, it's like, it doesn't matter if I'm wrong, as long as I've made a good faith prediction, and to the best of my model's ability, and great, that gives me information, now I can refine my model, and you know, what, what did it get right, what did it get wrong, um, and that's just how science evolves.

Milosz: 27:10

Absolutely. I think one interesting feature of single stranded nucleic acids, be it DNA or RNA, is that We have a lot of intuition about how proteins move right now from this aggregated mass of simulations But when you see a single stranded RNA doing something weird, it's hard even to know Oh, you know, is it completely wrong? Is it kind of slightly wrong or is it realistic? Because We know the constraints from the folded pieces that are there. But then as you say, the experiments often miss a whole part of the structure, for example, right? We don't know at all what it's doing.

Justin Lemkul: 27:49

Right, and that's, that's one of the areas we got into, I mean, even with the, the Drude RNA force field paper, We did a huge number of systems. Well, maybe not huge, but it was a lot for us. It was as many as we could really handle that had reasonable data available for them. And there were some times where it's like, well, that loop kind of behaves strangely, but I've got nothing that says that that's wrong. Um, so it's probably plausible behavior in the absence of any experimental contradiction. Um, you know, we, we note that as a possible area to investigate further. And, and we go with it, but you can only do so much. It's like we used all the data we could find for all the systems we wanted to do. And there were still things where, you know, at the end of the day, we were sitting, looking at the data going it, I guess it's reasonable, you know, and you hate to have that level of uncertainty but you know, it's a prediction, but let's see if it pans

Milosz: 28:38

Heh heh. Totally. I remember there was Almost a scandal when people showed the first simulation of proteins, right? I think in the 60s or 70s and people were like, oh, this is so floppy. This is so like jumpy while this were I don't know picosecond simulations, of course, but

Justin Lemkul: 28:53

Yeah.

Milosz: 28:54

were kind of scandalized by Right, so I think we're facing a similar question now with RNA, right? What are what is the real dynamics? Okay, Justin, uh, wonderful. Thank you for the great discussion. I think we made some very interesting points. Yeah, thanks for sharing the history and for taking the time to be on the podcast.

Justin Lemkul: 29:20

I really appreciate it. And I'm glad to join you and, look forward to seeing you in the future. Talking more about nucleic acid stuff.

Milosz: 29:26

Absolutely. Thanks a lot. Have a great day.

Justin Lemkul: 29:29

Yeah. You too.

29:30

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