Episode 10 - Ariane Nunes-Alves: Kinetics in drug design, molecular crowding, and the social life of a PI

In the tenth episode, Ariane Nunes-Alves and I talk about a kinetics-centric view of drug design, making the case that modeling kinetics in atomistic simulation is an important frontier that, despite clear biomedical relevance, is rarely explicitly addressed either in model parameterization or with the latest AI methods. We discuss the need to turn to a more explicit image of the drug's pathway towards its target, including not only affinities and residence times, but also the cellular environment. Then, we switch to the social sphere to discuss how much of a PI's professional time circles around interpersonal interactions, and where can we computational scientists go to find our community in the era of fragmentation and specialization of social media.

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, my guest is Ariane Nunes Alves, a group leader at Technische Universität Berlin, who is renowned for her work on methods and applications for the estimation of kinetics of ligand binding and unbinding. We often think about biology in terms of affinities and free energies. And of course, it's a convenient and well defined framework. But there are important problems in the field of drug design that require optimizing binding kinetics rather than affinity and it's an interesting challenge that cannot so far be addressed, for example, with structure based AI tools or even coarse grained simulations. So we start the conversation by asking how well the current atomistic models can even predict the rates of dynamical processes, given that our force fields are rarely explicitly parametrized. to realistically model such time dependent processes. We also discuss molecular crowding, another element that is usually missing from our models that nevertheless largely affects the behavior of molecules in a cell.

1:09

Then Ariane makes a point

Milosz: 1:12

that nowadays a significant portion of a PI's work consists of social interactions, be it with fellow computational scientists, collaborators, group members or students. And now this should be reflected in the focus of hiring committees. Finally, we wrap up the discussion by wondering if our community still has a social platform to coalesce around and whether that's something we could work on. Lots of questions ahead, so let's not wait any longer. Ariane Nunes-Alves welcome to the podcast.

Ariane Nunes-Alves: 2:00

Thank you for the invitation. Yes.

Milosz: 2:03

So Ariana, when I think about the drug design people, a lot of them seem to be thinking of molecular interactions in terms of affinity or you know, how strongly things stick together. And it kind of makes sense because it's such an easy target, such an easy objective to define for your drug design pipelines. But you're the one who exposed a kinetic centered view, right? So explicitly thinking in terms of the time it takes to bind or unbind. And so can you give us your take on why this is important? Why this question makes sense, perhaps what inspired you to look into it and how we can even start to approach it as computational people,

Ariane Nunes-Alves: 2:43

Oh, a lot of questions in one. So, yeah. First of all, why, uh, this is important? So, binding affinities, they are related to equilibrium properties, which is okay for in vitro experiments. But then when we think about the in vivo setting or about our cells, our cells are no equilibrium systems. They are out of equilibrium. So if you're equilibrium, we're dead. And then, uh, the best, way to describe, not the best way, but, uh, a more reasonable way of describing non equilibrium systems are kinetic parameters like, uh, association and dissociation rate constants. And then the other thing is that, it's usually, It's easier computationally speaking to study, equilibrium properties like, predicting, I don't know, dissociation constants, because, uh, we can predict them by only looking at the end states, the bound and unbound states. Time scales, they depend not only on the bound and unbound d states, but they also depend on the pathway we use to move from one to another. So this makes the problem way more computationally demanding. And I think that's why I went into this area, because I really like pathways for binding and unbinding. Yeah, I think it's cool.

Milosz: 3:56

I think it can be rewarding looking at all the movies that were produced of these processes, the question could be because my idea is that many of the models that we have were actually modeled to reproduce exactly the equilibrium properties, right? So what, confidence do we even have that the kinetics that we have in the physical simulations are realistic,

Ariane Nunes-Alves: 4:21

Ah, that's a fantastic question. Yeah, so, what I always try to do in my research at least, is to, reproduce experimental data. And in this sense, uh, we have, experimental, kinetic rates, uh, experimental dissociation or association rate constants for, Surface Plasmon Resonance or other experiments, what I mostly use nowadays is PowerEngine developed by my former, boss, my postdoc boss, Rebecca Wade, and in PowerEngine we can reproduce, Not the absolute values, but the relative kinetic rates for a group of compounds. And that's what gives us confidence we are the important biological detail there. But something that I think it would be very interesting to do is to come up with ways to verify that we actually have the correct pathways. I honestly, don't know a reasonable experiment, that can show us the pathway we see in the simulations corresponds to what we would see in

Milosz: 5:23

right? I can imagine that being done for like really, really large complexes where you can have a few probes. and the pathway is taking many nanometers, right? But not in a single binding pocket.

Ariane Nunes-Alves: 5:38

Yeah. Yeah,

Milosz: 5:40

equilibrium properties or time average properties of the system anyway. Yeah, that's, that's a great challenge. So yeah, how do we trust the results just based on what would be a single number, right? Dissociation rate, for example. Do you have any ideas for how to move forward with that? Or is it something that the experimentators will have to figure out?

Ariane Nunes-Alves: 6:03

I have no idea. Yeah. I guess, as you said, if you have a system with a longer tunnel, for instance, where we can maybe engineer metastable states, could think about designing some mutants that will Just deter the ground states and at the same time, stabilize, uh, another, intermediate state. Yeah,

Milosz: 6:24

so like design.

Ariane Nunes-Alves: 6:26

to do it, but yeah. It's hard

Milosz: 6:28

Okay.

Ariane Nunes-Alves: 6:28

convince experimentalists of my molecular dreams, but yeah.

Milosz: 6:33

That's definitely a lot of work. Yeah, more work than the actual scientific work ah, the convincing part. Yeah. So what for example, what are the biological cases where this actually may A difference, if you can bring up some projects in which, this question of optimizing kinetics made an impact in the biological problem.

Ariane Nunes-Alves: 6:54

Ah, sure. So, uh, there are a few papers that show, for instance, that for some specific systems people have seen that residence time, which is the inverse of the dissociation rate constant, is better correlated with in vivo efficacy, than equilibrium properties. That's one thing. There was this consortium, uh, before here in Europe between pharmaceutical companies and universities called, uh, K4DD. Uh, kinetics for drug discovery, they had also some examples, where increasing the residence time or increasing the time, the drug spends bound to a specific protein, They could have a stronger, effect for a specific drug, for instance, or, or in another case, they actually wanted to reduce the residence time because if the drug spends too much time bound to the protein, this would lead to undesirable side effects.

Milosz: 7:53

Right. So very case specific, yeah, that's what

Ariane Nunes-Alves: 7:55

hmm

Milosz: 7:56

the bottom line would be. Yeah. If you think from the drug's perspective, right, there is a moment where you enter the cell and perhaps at some point you bind to your target, and then you bind to something that tries to metabolize you, and that's when you're gone, right? So it really makes sense through this lens, how those properties can

Ariane Nunes-Alves: 8:16

Yeah, if you if you look at the cell environment, it's actually a very crowded system with a lot of different, macromolecules like nucleic acids, like a study of proteins. And then there is a lot of opportunity for off target binding or binding to an enzyme that will degrade you. So ideally It would be good for the drug to have a fast association rate constant towards its target protein because then it binds fast and then it can also offset the competing molecules that also want to bind there.

Milosz: 8:48

All right, that's a great point. Do we have ways to simulate that, Are we taking that into account?

Ariane Nunes-Alves: 8:54

Yeah, so I think the next step for, not only for protein ligand binding, but for biological phenomena in general is to go in the direction of simulating a more realistic, cellular environment. And this is something I want to do, but yeah.

Milosz: 9:09

do, we have even a template for that? Like, if someone wants to simulate a typical human cytoplasm, can you just take a data set of, you know, what would be the typical components and put them all together. That would be actually an interesting project if it doesn't exist yet.

Ariane Nunes-Alves: 9:26

Yeah, so I remember that in JCIM, uh, maybe three years ago, someone made, a proposal, made a publication there, to set up, the environment to simulate the cellular cytoplasm. And actually, nowadays, we do have, the information we need because we have a lot of proteomic data available. And then we know, Which proteins are available in the cell and the amount in which they are available. So in principle, the work is possible, but in practice, it's very challenging because you have to integrate a lot of, this peptidomic information with, structural information from the protein data bank or get models from alpha folds too. and also when you set up the system or the model, you have a very, very large system and then how do you simulate it? So people usually simulate it. Piece of the cytoplasm and not the whole thing. Yeah.

Milosz: 10:16

Would have to be a selection.

Ariane Nunes-Alves: 10:18

And I'm currently writing a review about it together with Rikiya Gresh. Uh, from New York, so yeah.

Milosz: 10:26

Amazing. Yeah, it would have to be some sort of selection of averaged property proteins, right? but it would be awesome if one day instead of putting just a box of water, you could put a box of cytoplasm inside your simulational model. Um,

Ariane Nunes-Alves: 10:40

will be very cool, Yeah

Milosz: 10:41

yeah. And see how all that affects what we thought is just a kind of silly equilibrium property or kinetic property of a system we're studying. Because I don't know if you know of many cases where exactly people try to apply this approach and it changed things drastically are you aware of such cases where, you know, where we saw the effect of environment in a simulation? Um, making a huge, huge, contribution to, the quantity of interest, whatever that was.

Ariane Nunes-Alves: 11:09

I have never seen such studies. No, I'm lying. there was

Milosz: 11:13

Hmm.

Ariane Nunes-Alves: 11:13

recent publication from Mihael Feig, Nature Communications, where they show they didn't actually simulate the whole cellular environment, but one way of simulating, of mimicking The crowded cellular environment is to use just one type of protein crowding a high amount. Then let's

Milosz: 11:34

All

Ariane Nunes-Alves: 11:35

the excluded volume and in some sense, also the weak interactions that proteins or the ligands you have with the Crowder, that in this case is just one type of protein. And then as a model Crowder, they use bovine serum albumin, they saw that, in the Crowder environment, the binding pathway, between an inhibitor and kinase changes. This is a very interesting result.

Milosz: 12:00

right. That sounds very, very cutting edge. My point is those questions are really fascinating from the point of view of the point that was made many times before on this podcast, which is we're trying to approach the biological, right? We are starting to get past the phase of reproducing in vitro environment where everything is controlled and maybe they're just PBS buffer and going to the wild west of the cell, right?

Ariane Nunes-Alves: 12:26

that's how I feel about it. The cell is currently the wild west. There are a lot of opportunities, and yeah.

Milosz: 12:32

Yeah, we had, we had a conversation with Pilar Cosio, for example, where she mentioned, cryo EM tomography. Okay. That actually shows you how crowded this whole thing is and people are trying to place proteins in their actual Positions discovered experimentally and I think that can be kind of synergistic with what you are describing, right?

Ariane Nunes-Alves: 12:53

Mm hmm. Yes.

Milosz: 12:56

many interesting frontiers there emerging these days Okay. Now from the, from the sciency to the social. So you pointed out and, uh, again, that's in line with other discussions we've had here how important the interpersonal skills are when evaluating potential hires for a faculty job, right? And we actually first met at a social event. So perhaps we've got this checked. But is scientific collaboration at the core of your argument or are you making a broader point? It goes beyond just scientists working together on various problems.

Ariane Nunes-Alves: 13:32

Oh no, uh, my major, concern when I made this point is that, uh, from my experience in Brazil and also here in Germany, see, so first of all, and now that I am a Junior Group Leader, I see that a huge part of the job is actually interacting with people. So, I always, I think I say to my students every week that this job of professor or group leader is a job for extroverted people because you really have to talk to a lot of people. And I'm not an extroverted people to be honest, but I mean, I think I can. Manage social interactions, but my main concern, uh, when I said this is that, I think, a major part of the job is teaching and supervision. So we interact a lot with people to teach them, and to mentor them. and I am kind of or upset the amount of stories I hear of, PhD students or master's students which were yelled at or treated poorly in whatever way. And I think this greatly comes from the fact that, when people are selected for this, group leader or professor positions, people just look at the scientific achievements and they don't really care whether people will be able to deal with and manage other people. And I think this is a very, part of the job because in the end of the day, we are here to serve and to teach, people, younger people. And then, yeah,

Milosz: 14:56

Right, it definitely doesn't get communicated in the stereotypical image of a professor, right? The stereotypical professor is usually kind of introverted. Although, as I also pointed out before, you know, people Solvay Conference or all the famous interactions between scientists. They were interacting. They were not that introverted. so even this public perception is, is kind of skewed in strange ways.

Ariane Nunes-Alves: 15:21

yeah, I think the public perception is the genius scientist alone in the corner writing equations or investigating something alone. But so far from reality. You really have to later on communicate your science, discuss with other people. So, yeah, this is another part interpersonal skills. So, yeah.

Milosz: 15:42

Right. And it's enough to look at a Twitter biophysical community to realize that people are actually quite extroverted and at least the people who self select to go there, right? I mean, maybe it doesn't apply universally.

Ariane Nunes-Alves: 15:54

Yes.

Milosz: 15:55

is definitely a bias But yeah, this whole question of community. I mean, I am trying with this podcast to create a feeling of community that, you know, we scientific people are trying to talk to each other and bring up, common topics and find common themes But, yeah. how do you envision the community? I mean, we, we have it now. So divided into different platforms into different subgroups Would you say there is such a thing as a the biophysical computational community?

Ariane Nunes-Alves: 16:24

That's a difficult question. I think we could be together a bit more, for sure. I when we go to conferences, for instance, you and I, we were in the European Biophysical Society last year. I think there is a friendly community at least among people in our generation. So, yeah.

Milosz: 16:44

Right

Ariane Nunes-Alves: 16:45

but I think

Milosz: 16:45

I see.

Ariane Nunes-Alves: 16:46

for instance, is a great idea. People were a bit more active on Twitter before, but now with Elon Musk, I feel like things were diluted a little bit, and we still haven't found the place to get together again.

Milosz: 16:59

That's, yeah, that's part of my concern that some people left for Mastodon, some people stay on Twitter, some people left social media for good. And, it was my, my general question. Where do scientists meet if they don't, for example, go to a conference at a given moment, right? What is the new meeting place?

Ariane Nunes-Alves: 17:17

Mastodon, and I found it hard to find new people. Uh, I tried Blue Sky, and it didn't catch up with me yet. PhD students are younger than me, they talking a lot about LinkedIn. But I've seen it a couple of times and I don't really like the self promoting vibes, to be honest, but I may end up going to LinkedIn, I don't know.

Milosz: 17:40

I think it's hyper professionalized. So I don't really see LinkedIn as a replacement for Twitter in that aspect, right? would be closer, but Reddit is also for much younger folks, I feel. Uh, so I already feel a bit,

Ariane Nunes-Alves: 17:53

I've never used it,

Milosz: 17:54

Too old, but I know that there are, for example, by informatics communities, but they're mostly for students, people looking for advice and, more like undergraduates than, the professional level. Yeah, it's a really tough question. I think I'm one we could be paying more attention to. I don't know if anyone has any good ideas in that regard, but I really feel that the demise of Twitter was, uh, bad moment for the biophysical community. But then, yeah, it's really interesting what you say about the ultra or the hyper social aspect of, the life of a professor. Do you think it's, also depending on the culture or it's something that's universal? I mean, you have background from quite a few institutions. So,

Ariane Nunes-Alves: 18:40

I think that's That's the professor life, yeah. I don't think this would change a lot across different countries, I think, that's the way things are probably supposed to be,

Milosz: 18:51

all right. So wonderful. Ariana Nunes Alves. Uh, thank you for the illuminating discussion and for being on the podcast.

Ariane Nunes-Alves: 18:59

Thank you for the

Milosz: 19:00

Uh,

Ariane Nunes-Alves: 19:01

Pleasure. Yeah,

Milosz: 19:02

it was great. Hope you have a great day.

Ariane Nunes-Alves: 19:05

you too.

19:05

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