 Hello and welcome to the OIST podcast, bringing you the latest in science and tech from the Okinawa Institute of Science and Technology Graduate University. My name is Lucy Dickey. In this episode, I speak to Professor Greg Stevens about his research into the physics of behavior and why it's important. Have a listen to this excerpt. Movement is really the output. It's the place at which evolution acts. You know, whether you are successful finding food or finding mates comes down to how you move. And so we think that that's a really good place to look if you're going to find unifying principles in biology. In a nutshell, Professor Stevens is a physicist who looks at biological questions. He's an adjunct professor at OIST and his group is focused on quantifying behavior. You might be wondering their behavior of what and the answer is pretty much anything from single-celled organisms to crowds of people. His group has previously looked at how worms wiggle, whether machines can be taught to automatically track every bnr beehive, or if the movement of zebrafish can be measured. When we spoke, he was in semi-lockdown in Amsterdam and I was in semi-lockdown in Okinawa. So this interview was the first episode of the OIST podcast that was conducted via Zoom. It came about because Professor Stevens had recently co-organized an online workshop on the physics of behavior, which attracted nearly 1500 people. Given how popular this kind of research seems to be to an audience of both other scientists and the interested public, I decided I wanted to find out more. So my name is Greg, Greg Stevens. I'm an adjunct professor at OIST and an associate professor at the Frey Universitat in Amsterdam. My research specialty is theoretical biophysics and our group focuses on behavior. So instead of thinking about neurons and genes and molecules, the small-scale biological things that make up you, we think about you. And how does this all work together? How does this all come together? So could you go into a bit more detail about what you mean by the physics of behavior? Sure. And so, you know, physics of behavior is, I'll say it's kind of a grab all sort of term. And what I mean by that is traditionally ethologists, sort of biologists with specializing in thinking about behavior would maybe go into the woods or go into the water in Okinawa and dive into the water and watch the fishes or other animals and watch them in their natural habitat, maybe take notes on what were the important behaviors. And there's a lot to learn from that field of ethology. But in a physics of behavior approach, we try to take selected systems into the lab. We build new measuring experiments. We have one at OIST where we're watching, we built a 3D tracking system for multiple fish. We watch these, in this case, these pairs of male fish interact with each other under conditions where we can really sort of follow exactly what their bodies are doing in time. And so, we build these, so that we have these new measurements, they're quantitative, and they're high resolution in space and time. And out of those measurements, we try to ask independently of what your eye sees as important behaviors. How can I use sort of statistical mathematical techniques to quantify the behavior that we see? So physics of behavior really means sort of quantifying in a more objective way the behaviors that we see and ultimately looking for why do we see the behaviors that we see? What are the underlying principles that govern how you move? That's really interesting. So what first inspired you to look at this field? Right. So you have to go back to, I was a postdoc and I knew that quantum gravity was, didn't have this richness of this data experience, so I was looking for the connection between theory and experiment. And also, I was still a little bit, let's say, under the sway of these physics ideas about what is fundamental, right? The idea that, you know, quantum mechanics and general relativity, that's what's fundamental. But then, I was a postdoc at Los Alamos, a big research laboratory in the United States. And I was lucky enough to be hired as a postdoc with a small group of physicists thinking about the brain. And once you start thinking about the brain, then you start to think, huh, I wonder whether it's, what fundamental means is that matching quantum mechanics to general relativity? Or is that understanding the brain that matched quantum mechanics to general relativity? That is, we have this tremendous capability capacity as humans to build science, right? And the laws that we build, especially with my experience in physics, they are remarkably powerful. We can reach out to the smallest scale sub, you know, within atoms or the largest scales of over the universe and have predictive, we have predictive power. And that's pretty crazy, right? And somehow our brains do this, even though we don't have direct sensory experience. So that was a long-winded way of saying that, okay, well, neuroscience is really interesting. There's lots of room. So these, there's many biological fields that are welcoming to people like me, to physicists with good quantitative background, thinking about, thinking about their problems. And I was working on this in Los Alamos, and then I got an opportunity to be a postdoc at Princeton with Bill Bialik, who's a strong mentor for many in theoretical biophysics, sort of provide the welcoming home for physicists who are learning how to think about biology, but still maintaining their identity as a physicist. I think that was, for me, that was a, that was a fundamental, that was a fundamentally important sort of mark in my career. And when I arrived at Bill's, in Bill's group, this is kind of gives you a sense of the practical ways that one chooses directions. There were lots of postdocs thinking about the brain and thinking about, as a physicist, how do you think about the brain? How do you think about neurons? And, you know, in a brain such as yours, you have 100 billion neurons, a huge number, about the same number as there are stars in the galaxy, just an astronomical, literally an astronomical number of neurons. And typically you measure only at best a few thousand. So you're, you're measuring a small fraction of what's relevant. And I was worried that with, with so with, with by only measuring such a small fraction, how could I possibly learn the computations that the brain is doing? And, you know, if you put that, that worry alongside the fact that everybody was working on neuroscience. And then there was this one project with this new, new faculty member at Princeton on thinking about this wiggling little worm. And I was like, well, that sounds cool. How do you quantify how the worm wiggles? What a crazy sounding idea that no one had really, no one really had thought about that before from the physics perspective. So it took a while. We had to learn a lot of new stuff. We're still learning a lot of new stuff, but it was kind of that was sort of the beginning. And then we, and then this blossom at the same time that I started doing this, really, there was no field of physics of behavior. So we started reaching out, pulling in, pulling in sort of ethologists, other biologists, neuroscientists to sort of build up this quantitative representation of behavior. Now I think it's really a field. There's a lot of people who are interested in there's a community to talk to. And it's really quite something. Yeah. Cool. So I read on your website that the physics of behavior looks from individual organisms to entire societies. And I guess you've talked about how you looked at the worm and how how it behaves and trying to quantify that. But could you give an example of what you mean by entire societies? Sure. So we had, she just left our group. We had a wonderful postdoc in our group, Keisha Bozak. And she came to join, she joined OIST with a kind of crazy idea, which is could she teach machines to automatically track every bee in a beehive? And from my perspective, this was great because if you could follow every, typically there's a thousand or more bees in a beehive there, you can arrange, you can, you can build a home for them so that the bees live in a quasi two-dimensional plane. So it's amenable to imaging. And if we could track every bee, then it would be like, then we would sort of have, it would be like a bee gas. And we'd have to figure out what kind of gas this was and what that all meant. And so Keisha then spent the last few years building this capability. It's remarkable. She just put a paper on the archive from this work. So we're, we're getting feedback on that paper now. And what I mean by tracking an entire society is we are watching every bee in this hive, sometimes over long time scales, like months, sometimes over short time scales, and we can sort of follow all these patterns and we can ask, well, if you're, when you're a young bee, what kind of task do you do when you're an older bee? What kind of task do you do? What are the collective behaviors of bees? So how do bees work together? Maybe they want to heat up the nest or cool the nest down? All these kinds of questions. It's kind of like you have access to all the people in the city and then you have to figure out how the city works. That's what we're doing with bees. And what do you think is next for that project? Well, so it, if you think about the next, next for that project, there aren't really a lot of good ideas yet for, if you measure all the individuals as they're moving around, especially when they're moving around very dynamically, what, what are their collective, what are their collective modes? And our next goal in that project is to actually learn about these collective modes. And it is very similar, conceptually to the early days of thermodynamics in physics where you have a bunch of gas molecules like the gas in this room, you have a huge number of molecules. And yet if you're going to move this gas through a process and make a steam engine and do work, you know that you only need to pay attention to a small number of those variables like the pressure and the volume of the temperature. So can we find those kinds of variables for the bee colony? They're going to be different. It's not like it's a gas, it's not a physical gas. But maybe we can find the right collective variables that give us some understanding of how the colonies actually control. Earlier on in this interview, you talked a little bit about studying worms. Could you expand a bit about, a bit on this? When we first started studying worms, and this is, this is the new, for those that are experts, this is the nematode C. elegans. It's interesting for historical, the history of OIST too, because Sydney Brenner was the first president of OIST and he was really the biologist that brought C. elegans as a model system into biology. And it's one of the smallest brains in the world. It wiggles around in the soil and finds, finds mates and food and has predators to escape from and is a kind of a model system in many ways. And yet when people would study its behavior before we got started, more often than not, that would mean putting a graduate student in front of a camera and having them describe the behavior sort of by eye. And our first goal then is, well, we thought we should be able to describe quantitatively how the worm is moving. We ended up doing that and inventing a new term sort of without, without planning to. So there's a term that's now quite, quite common called eigenworms. And eigenworms are a low dimensional space of postures that we found at the worm and habits. And so instead of having to sit someone in front of an image or a camera and describe the behavior, we now have this sort of quantitative framework in which there's a five dimensional time series that, that, that we analyze and that's what the worms behavior turns out to be for us. With all this talk about quantifying behavior from worms to fish to honeybees, I asked Professor Stevens if these techniques could be applied to humans and human society as well. He responded that they absolutely could and people did. In fact, there are a lot of similarities between their behavior of even the tiny wiggling worm and humans. So we work on systems that are, that are a little bit easier for us to measure in the laboratory. But there's a lot of commonality between what we find in, even in, in the, in the little worms and what we might find in you. So for example, the worms are, they have about 70 some muscle cells, you can count them exactly, 72, I think, if I remember correctly. And, and yet the number of postures that they actually use is a much smaller dimensional space. That's the space of eigenworms. The same thing is true with you. So if I watched all of your, all of your joint angles as you move around and as you move your, your hands, then I might say, well, you're, you're, you have all these degrees of freedom. You could be super complicated, but it turns out that you use a much smaller degree set of, set of motions and we can find those set of motions. So if I really did use one of these tracking algorithms to follow your joint angles, I would find just like we found with the worms that you, that your posture is described in this kind of lower dimensional characteristic set of motions. And we haven't studied humans in yet, but it's certainly something that we could do. And whether it's using cameras like the Kinect system or other ways of following, I think it would be quite interesting. And even, you know, we study, as you mentioned, we study these interacting fish and they do some sort of a colloquially, they do some sort of a dance when they're interacting with each other. I think it'd be really cool to watch people as they do, as they do a dance and in the posture of the way the posture and the, and the changes in posture between you and your dance partner, I think we learn a lot about what's going on in the brain. So it can be applied to pretty much anything across like the biological world. It can be applied to pretty much anything in terms of behavior. So the challenge is, the first challenge is, can you measure what you think is relevant? So if it's changes in posture for you, I have to be able to follow, you know, it's not just about where you are in the world. I want to be able to measure what your posture is. Same thing for animals. Some animals are more challenging than others. For example, if we look, we have a new collaboration starting with, with Sam Ritter at OIST and on cephalopods. And if you look at all the degrees of freedom that a squid can engage in, well, that's pretty complicated. So it's going to take a while before we learn how to measure them. But that's in principle, you could think about different, all different kinds of animals, including even beyond animals down to single cells and how they move and everything. I think, you know, the connecting, what connects all of these systems is, this is really the output. Movement is really the output. It's the place at which evolution acts. You know, whether you are successful finding food or finding mates comes down to how you move. And so we think that that's a really good place to look if you're going to find unifying principles in biology. How do you foresee this being used in the future? So if you want a finer scale measurement of behavior, because you want to ask, well, how does this gene say, or how does this, how do, how do these sets of genes control my behavior? Then this gives you a much finer, a finer tooth comb in order to look for, for that, look for that. If you want, if you want to find the effect of a drug on behavior, this gives you, you know, you can do, you can do screens in a high throughput way because no human has to be involved. These are sort of practical, practical concerns you could predict. You mentioned humans, you know, we think of ourselves as, as guided by whatever whims are in our brains at the moment. But you also know that you have strong predictive patterns in your life. And I think we could learn some of that, quite a bit of that if you watch people over time. But for us, you know, as a, as a group sort of really focused on, on, on physics, the question is really, can we very much like thermodynamics or even, even like fluid dynamics, can we build a description at the level of behavior that isn't as complicated as all the genes and all the neurons that make it up. But it's some sort of effective description like pressure or volume, but that still gives us strong insight into how the animals are actually behaving. So for me, it's really a chance to, to write down on the scale of behavior, a theory that, that would be predictive and that would really lead to insight into how, how animals actually move. Cool. So you've touched on a few points with this next question, but I was wondering if you could expand on what applications does this have for society? Let's see, big questions. You could track your cat, and people do. So there, there are ramifications for society as a whole that we should be having discussions about and it's already happening. So there's already in different, in different societies around the world, there can be sort of mass surveillance, which includes sort of, you know, can go all the way down to the level of predicting behaviors. Maybe you, maybe you want to understand whether this person is, is about to do something. And I think that's coming. It's here. And it's something we should talk about on the level of humans. So if you want to understand how a drug works, you want to look for, you want to look for a drug that does something, you need to understand how it impacts behavior. So high throughput screens and model systems, like the worms, like Drosophila, but also including animals, you know, closer to humans, mammals, like mice and fish. So I think in terms of, in terms of doing high throughput anything on, you know, behavior is, is one of the signals in biology that really matters. So you do work on a lot of different projects. Do you have a favorite project, one that you've enjoyed working on the most so far? So they all tell us something kind of different. C. elegans, the wiggling worm is, is the system for which we have the most experience. It's also the system for which we have, I think, the most sophisticated theoretical understanding. We have the most data for that system too. But there are things that a little wiggling worm doesn't do, right? It doesn't form these big societies like a honey bee. It doesn't have these social interactions. The zebra fish are remarkable for how social they really are. We put, we put two fish in the tank and they just, they really like to hang out with each other. And so each one of the systems, you know, in this overarching theme of physics and behavior, each one of these systems really tells us something different. So it's, it's, we don't really have favorites. They're all, they're all our kids. Cool. And finally, to finish up, are there any potential future projects that you're really looking forward to? Well, we're totally excited about, about Sam joining OIST and cephalopods are, are just crazy cool. And just to put it scientifically. So anyone who's watched an octopus or squid or other cephalopods, the cuttlefish, they are, they are intelligent. They have a lot of interesting behaviors. They communicate with their skin. They have tremendous emotion with their tentacles. And this is a, this is a crazy new direction for us. And so we're really interested to see what we can learn about such a crazy animal. I think that's, that's certainly one excitement that's on the horizon. Working with cephalopods certainly does sound very cool. And I'm looking forward to seeing where the research goes within this field. In the meantime, thanks for listening to the podcast. It was recorded and edited by me, Lucy Dickey. Special thanks to Professor Greg Stevens for his time and for explaining his research so clearly. If you enjoyed this episode, subscribe so you can hear more as soon as we release them. You can also find us on Facebook and Twitter. Or send us an email to media at oise.jp. See you next time.