WEBVTT 00:00:00.017 --> 00:00:04.977 So Paul Voucher and Tony Prescott. And Partha was describing to us an amazing 00:00:04.977 --> 00:00:13.437 project where you have this sort of massive automated data collection on a very 00:00:13.437 --> 00:00:19.677 detailed anatomical level that can lay a foundation for new insights in the brain. 00:00:19.837 --> 00:00:24.317 So Partha, why don't you try to give us a short summary how you ended up doing this? 00:00:25.117 --> 00:00:33.937 I was really shocked to find out how little we know about how the brain is wired. 00:00:34.137 --> 00:00:35.617 That's what got me started. 00:00:35.977 --> 00:00:40.297 I trained as a theoretical physicist, and when I came into biology, 00:00:40.497 --> 00:00:49.237 I'm slowly coming to grips with the knowledge that we are very much limited by lack of information. 00:00:49.537 --> 00:00:53.537 And we have this paradox that we are data rich and data poor at the same time. 00:00:53.537 --> 00:00:57.497 We have lots of data, but sometimes critical pieces are missing. 00:01:00.637 --> 00:01:07.477 And once I realized that even in the most thoroughly studied model organism, which is the rat, 00:01:09.117 --> 00:01:14.637 the wiring diagram of the brain at a scale that neuroanatomists have studied 00:01:14.637 --> 00:01:22.877 has large empirical gaps in it, I thought I would set out to try to rectify that situation. 00:01:23.697 --> 00:01:27.797 But in some sense, it's very counterintuitive because often in neuroscience, 00:01:28.197 --> 00:01:31.857 also when you talk with other neuroscientists, it's this feeling like, 00:01:31.897 --> 00:01:33.197 well, we're drowning in data. 00:01:33.277 --> 00:01:38.557 We have so much data, different scales, different species, physiology, anatomy, etc. 00:01:39.397 --> 00:01:42.817 So how do you position that with respect to this data richness? 00:01:43.037 --> 00:01:44.997 So how do you now deal with this paradox? 00:01:46.977 --> 00:01:51.697 I think the way I think about it is that we are data rich. 00:01:51.737 --> 00:01:55.717 You're quite right. there's more than half a million articles being 00:01:55.717 --> 00:02:00.317 published on abstracts being published on PubMed every year so there's a tremendous 00:02:00.317 --> 00:02:05.597 amount of information out there I think this information is very heterogeneous 00:02:05.597 --> 00:02:07.877 and not really integrated there 00:02:07.877 --> 00:02:13.157 is no super brain that really has all the information and so does you, 00:02:14.857 --> 00:02:21.517 so yes we are data rich in this very heterogeneous sense lots of bits of data. 00:02:21.857 --> 00:02:27.497 We're data poor in the sense in neuroscience compared with, let us say, 00:02:27.537 --> 00:02:32.857 in genomics, where we have comprehensive information about a certain scale of organization. 00:02:33.297 --> 00:02:36.377 Let's say the genome. We've got the full genome of multiple species, 00:02:37.417 --> 00:02:43.037 but we can't say the same at any scale really of any organism for the nervous 00:02:43.037 --> 00:02:47.997 system, that we have the full set of information with the exception of C. 00:02:48.037 --> 00:02:54.757 Elegans, which has been kind of the you know example that people have put out um so for me the, 00:02:56.132 --> 00:03:00.292 Reconciliation is that on the one hand, we do have a lot of heterogeneous, 00:03:01.652 --> 00:03:06.412 unintegrated information, sometimes partial, because people have focused, 00:03:06.792 --> 00:03:09.652 over-specialized on certain areas of the system. 00:03:09.812 --> 00:03:12.852 A lot of studies have been attempted, a lot of studies of the visual system. 00:03:15.752 --> 00:03:24.892 So if one takes an integrated perspective and does a project where you map out, 00:03:24.952 --> 00:03:30.032 as in our in our case, the whole brain circuit at some level of resolution. 00:03:31.912 --> 00:03:35.572 Yes, that's generating a significant data set. 00:03:37.572 --> 00:03:44.132 But at the same time, it's not even at the scale of this large heterogeneous data that's out there. 00:03:44.472 --> 00:03:50.472 But given how you summarize this argument, you're facing two challenges in some sense, no? 00:03:50.552 --> 00:03:53.512 Because on the one hand, isn't one implication that you say, 00:03:53.672 --> 00:03:57.752 well, yes, we have all this data, But you might as well have not collected it 00:03:57.752 --> 00:04:00.272 because it's not organized in the right way. 00:04:00.392 --> 00:04:02.592 Is that really the consequence of what you're saying? Would you agree? 00:04:02.952 --> 00:04:07.192 Yeah, I mean, I think that that's one way of putting it. I think a good analogy 00:04:07.192 --> 00:04:10.532 that is certainly not new at this point is thinking about Google Earth. 00:04:11.952 --> 00:04:15.292 We probably have a lot of information about geography, about, 00:04:15.412 --> 00:04:18.132 you know, restaurants in a particular city and so on and so forth. 00:04:18.512 --> 00:04:23.492 But until there was a framework on which to hang all these pieces of information, 00:04:23.752 --> 00:04:25.212 they remained disparate. it. 00:04:25.232 --> 00:04:31.272 So perhaps this project that we've initiated, one thing that it will do is to 00:04:31.272 --> 00:04:35.152 provide a framework in which to collect this heterogeneous information. 00:04:35.552 --> 00:04:39.272 If that succeeds, then, you know, we will resolve the paradox to some extent. 00:04:39.872 --> 00:04:43.092 Underlying your argument, though, is another assumption, which is really interesting, 00:04:43.252 --> 00:04:46.852 because sometimes you're saying, well, the human genome project was actually 00:04:46.852 --> 00:04:50.572 a great success, and I want to follow this paradigm now from neuroscience. 00:04:51.152 --> 00:04:56.272 But is that really a reasonable assumption? I wouldn't quite put it in those terms. 00:04:56.452 --> 00:05:04.592 I would say that, yes, it was a success, but success as defined in certain ways. 00:05:04.852 --> 00:05:09.812 I think one way in which it succeeded is providing this integrated framework 00:05:09.812 --> 00:05:11.392 for pulling information together. 00:05:11.692 --> 00:05:19.212 Another way in which it succeeded was it helped us gain understanding of the evolution of genomes. 00:05:20.112 --> 00:05:24.632 We've also gained information about human history. 00:05:24.852 --> 00:05:30.912 So, we haven't solved certain problems which were promised and maybe we will 00:05:30.912 --> 00:05:34.372 still solve those problems in the future, but in certain ways it has been successful. 00:05:34.772 --> 00:05:42.132 And I think in those ways is where I would want to borrow from the success or hope for the success. 00:05:43.691 --> 00:05:50.671 Looking ahead to when you've built this new huge data set, what's your idea 00:05:50.671 --> 00:05:52.191 about how people are going to use it? 00:05:52.271 --> 00:05:55.431 I mean, I can see sort of neuroanatomists looking at the raw data and say, 00:05:55.491 --> 00:05:57.411 yeah, this fills the gap. 00:05:57.691 --> 00:06:01.691 I don't any longer need to do that experiment because I can just look at your database. 00:06:02.031 --> 00:06:06.611 But for those people that want to draw something out of it to condense it down 00:06:06.611 --> 00:06:10.631 into some more compact understanding of the brain, what kind of tools do you 00:06:10.631 --> 00:06:13.031 think they will be able to use on this database? 00:06:13.691 --> 00:06:19.891 Well, we will try to provide integrated versions of the data at lower resolutions. 00:06:20.791 --> 00:06:25.251 So in the ideal world, what we should be able to provide is a probabilistic 00:06:25.251 --> 00:06:33.651 math saying that if you start from point A in a reference brain and you were 00:06:33.651 --> 00:06:35.871 a neuron with a cell body near point A, 00:06:35.911 --> 00:06:43.091 what's the probability that you had a projection or an arbor at point B? 00:06:44.431 --> 00:06:47.991 So, that sort of information integrated across the injections, 00:06:48.471 --> 00:06:50.311 we will be able to provide. 00:06:53.191 --> 00:06:58.991 That will then presumably help people who are trying to gain an understanding 00:06:58.991 --> 00:07:02.791 of the architecture of the whole brain circuit. 00:07:03.111 --> 00:07:04.251 I mean, I know that this is a 00:07:04.251 --> 00:07:07.911 nascent field in the sense that not a lot of people are trying to do that. 00:07:07.971 --> 00:07:13.591 There's more interested in understanding the microcircuits in a particular region of the brain. 00:07:14.451 --> 00:07:21.151 But we do hope to provide some integrated pieces of information apart from guiding, 00:07:21.331 --> 00:07:24.591 thinking about how the system is organized and how it works. 00:07:24.891 --> 00:07:30.651 My hope is that we will also get multiple species, so we will be able to address 00:07:30.651 --> 00:07:31.531 evolutionary questions. 00:07:32.491 --> 00:07:36.871 Previously, brain evolution has been studied largely in the context of shapes and sizes. 00:07:37.411 --> 00:07:43.711 So cortex has grown in size, the olfactory bulb has shrunk its size, and so on. 00:07:43.791 --> 00:07:49.251 But we haven't really studied it at the level of what circuits are and how they 00:07:49.251 --> 00:07:51.251 have changed, with exceptions. 00:07:51.591 --> 00:07:57.751 But it has been very much driven by non-circuit considerations, 00:07:58.211 --> 00:07:59.591 the evolution. I don't know if you agree with, 00:08:00.382 --> 00:08:05.002 Well, what I was wondering in that in that respect is also the incredible individual 00:08:05.002 --> 00:08:06.602 variability you might try to get. 00:08:06.762 --> 00:08:11.582 So how are you going to handle that? Right. So you might take, 00:08:11.642 --> 00:08:16.982 let's say, a mouse or a rat or or a monkey from which you get your anatomical data. 00:08:17.142 --> 00:08:22.522 But now you know that over individuals, there'll be an incredible dispersion 00:08:22.522 --> 00:08:24.382 of these kinds of projections. 00:08:24.662 --> 00:08:28.982 So how do you see yourself extracting the rules from that kind of variability? 00:08:30.382 --> 00:08:37.842 I do have a premise that the variability will not be so large that we cannot 00:08:37.842 --> 00:08:41.602 extract these regularities. 00:08:41.762 --> 00:08:45.842 So going back to a comment you made during my presentation, is that I do have 00:08:45.842 --> 00:08:48.422 a hypothesis, in some sense a weak hypothesis, 00:08:48.702 --> 00:08:53.642 that there is something to be learned from these experiments because there is 00:08:53.642 --> 00:08:58.082 some underlying organization that is relatively concerned from individual to 00:08:58.082 --> 00:08:59.762 individual in this mouse strain, 00:09:00.122 --> 00:09:08.342 we will do multiple injections and multiple repeats of the same injections in the same brain region. 00:09:08.502 --> 00:09:12.802 So empirically, we will hope to answer this question. I don't think it's a theoretical question. 00:09:13.042 --> 00:09:17.882 So I do have a theoretical hypothesis that the variation will be controlled 00:09:17.882 --> 00:09:22.922 enough that there is a meaningful mesocircuit, but we will try to prove it empirically. 00:09:23.302 --> 00:09:25.522 And how many different... 00:09:26.102 --> 00:09:29.042 Since the brain is continuously changing also ontogenetically, 00:09:29.902 --> 00:09:33.822 so how many measurement points do you think you would need in a single species? 00:09:34.542 --> 00:09:36.962 We are doing it in the mouse at a single age. 00:09:38.342 --> 00:09:42.222 We have not yet even started thinking about a developmental version of this. 00:09:42.622 --> 00:09:45.982 One thing that I'm hoping is that yes, 00:09:46.042 --> 00:09:52.822 we are running this particular experimental project in my lab but we are hoping 00:09:52.822 --> 00:09:57.862 to demonstrate that this kind of project can be done quite economically in multiple laboratories. 00:09:58.582 --> 00:10:05.142 So once that comes about, I think people should be able to fill in multiple ontogenetic steps. 00:10:05.542 --> 00:10:09.942 What we are hoping to do is to take one age and then look at different mouse 00:10:09.942 --> 00:10:14.782 strains, so different genetic constructs, and see how the circuit changed at 00:10:14.782 --> 00:10:17.162 a given age. Right. So how, 00:10:18.260 --> 00:10:20.840 There's something interesting about the trajectory that you went through. 00:10:20.960 --> 00:10:25.540 As you said earlier, you started as a theoretical physicist and then sort of 00:10:25.540 --> 00:10:31.160 discovering the brain, if you want, and in some sense now becoming much more 00:10:31.160 --> 00:10:32.560 an experimentalist in some sense. 00:10:32.740 --> 00:10:38.740 But then, let's say, a large-scale anatomist or industrial-scale anatomist. 00:10:38.740 --> 00:10:44.300 And if I look at the way you present the argument, it seems also interesting 00:10:44.300 --> 00:10:48.620 that along the way, you've been shedding more and more of your theoretical skin, 00:10:48.780 --> 00:10:51.200 right? I do it at night privately. 00:10:53.120 --> 00:10:57.960 Is that fair to say? No, my bedtime reading recently has been, 00:10:57.980 --> 00:11:00.860 you know, formal logic and models of real computation. 00:11:01.240 --> 00:11:07.160 I mean, I haven't shed my theoretical interests. It's just that I have become 00:11:07.160 --> 00:11:11.760 much more conservative than when I started in making theoretical models because 00:11:11.760 --> 00:11:15.980 I've come to realize how difficult the problems really are that we are trying to describe. 00:11:17.200 --> 00:11:23.320 What happened to me when I was a theoretical physicist is I developed a great admiration of theory. 00:11:23.580 --> 00:11:27.160 So I don't want to take the name lightly, so to speak. 00:11:27.600 --> 00:11:32.460 It's very interesting. So in some sense, what you're saying is that in this 00:11:32.460 --> 00:11:36.100 move from, let's say, the naive but maybe very ambitious theoretical person 00:11:36.100 --> 00:11:39.180 into now, let's say, the anatomist at the bench, 00:11:39.340 --> 00:11:43.160 you have acquired more humility towards the problem. Absolutely. Absolutely. 00:11:43.460 --> 00:11:46.620 That's the point. Absolutely, yeah. I think that these are very hard problems, 00:11:46.700 --> 00:11:52.740 and I still hope to make some theoretical contributions, but I would like them 00:11:52.740 --> 00:11:53.880 to be solid contributions. 00:11:55.044 --> 00:11:59.044 So where do you think we are in terms of theories of the brain? 00:11:59.204 --> 00:12:04.624 What aspects of the brain do we understand well, and where do we need to pay attention to? 00:12:05.044 --> 00:12:09.044 I think single-neuron biophysics is a clear success story, 00:12:09.764 --> 00:12:14.964 to the extent that there's a well-developed theoretical framework, 00:12:15.184 --> 00:12:18.884 and that it's got an attached experimental framework, and there's a close feedback 00:12:18.884 --> 00:12:20.924 loop between the theory and the experiment. 00:12:22.064 --> 00:12:26.824 Some other parts, I think it's piecemeal. I think certain systems in the brain 00:12:26.824 --> 00:12:29.224 are well-developed theoretically. 00:12:29.384 --> 00:12:34.204 I'm not saying that they are fully developed, but the sensory systems are relatively 00:12:34.204 --> 00:12:35.844 well-developed in that way. 00:12:35.904 --> 00:12:40.044 We have a good, at least, approach to the visual system or the auditory system. 00:12:40.804 --> 00:12:45.584 Also, to a lesser extent, but also to motor systems. 00:12:46.984 --> 00:12:51.804 At the peripheral levels, I think we've got good models. models, 00:12:51.904 --> 00:12:59.124 what I feel is lacking is a more systems-level understanding of the whole system. 00:12:59.484 --> 00:13:06.544 I think that the theoretical work that has been done there is non-mathematical, 00:13:06.544 --> 00:13:08.744 but still legitimate theorizing, 00:13:08.904 --> 00:13:14.324 and people who have done that are perhaps not thought of as theorists because 00:13:14.324 --> 00:13:17.604 they've really thought about the system, they've articulated their thoughts. 00:13:17.764 --> 00:13:22.104 I would still call that theory. I mean, Darwin didn't have a single equation, right? 00:13:22.364 --> 00:13:26.964 So what is the status? 00:13:27.344 --> 00:13:30.984 I think that there is a lot of work to be done. 00:13:32.004 --> 00:13:36.304 Progress seems to me to be pragmatically occurring in fields which are tied into robotics. 00:13:37.844 --> 00:13:40.884 Insect locomotion is one area that I've followed to some extent, 00:13:41.044 --> 00:13:42.884 and it seems to me to be working quite well. 00:13:44.853 --> 00:13:47.733 Of course, that's limited in the sense of understanding the brain. 00:13:50.133 --> 00:13:55.573 One lesson that I've drawn from there is really to think about engineering theories 00:13:55.573 --> 00:13:58.093 and to think about what the system does, 00:13:58.373 --> 00:14:02.553 what's the logic that one would employ in constructing systems like that, 00:14:02.653 --> 00:14:10.353 and then trying to understand whether the evolutionary processes are compatible with such thinking? 00:14:10.773 --> 00:14:12.873 Is there a role for convergent evolution? 00:14:13.473 --> 00:14:17.333 So if I were to say in a single sentence where I see the future of theory, 00:14:17.513 --> 00:14:20.153 I would say take engineering much more seriously. 00:14:20.413 --> 00:14:25.613 I think that it's certainly been a theme that has been explored since the times of cybernetics. 00:14:26.393 --> 00:14:31.613 It has perhaps been sidelined in some areas of neuroscientific theorizing, 00:14:31.733 --> 00:14:37.193 but I think it needs to be brought more to the fore and more training is necessary 00:14:37.193 --> 00:14:38.873 in engineering disciplines of the, 00:14:40.033 --> 00:14:41.773 some of the neuroscientific theorists. 00:14:41.953 --> 00:14:47.673 Are you saying that we should give up this more physics-based dream of, 00:14:47.693 --> 00:14:52.853 let's say, closed-form solutions of the phenomena that we're trying to study and go more towards, 00:14:53.173 --> 00:14:59.673 let's say, open solutions or descriptions of these systems? Is that a key step there? 00:14:59.953 --> 00:15:03.813 Yeah, I think that there has been a certain limitation in the theory community, 00:15:04.053 --> 00:15:11.933 at least those of us who came from a physics background to take a point of view 00:15:11.933 --> 00:15:16.113 that resembled too closely the successes of theoretical physics and simply hoping 00:15:16.113 --> 00:15:18.133 that the same trick will work again. 00:15:19.813 --> 00:15:23.753 I think that that's not working, is really the evidence. 00:15:23.973 --> 00:15:27.333 There are some isolated examples like the Hauss-Gnostic equation, 00:15:27.793 --> 00:15:31.293 a partial differential equation describing how the action potential works, 00:15:31.453 --> 00:15:35.433 might lead us to believe that in general, ordinary and partial differential 00:15:35.433 --> 00:15:40.933 equations are what we are looking for in terms of theoretical constructs. 00:15:43.427 --> 00:15:48.587 Again, looking in a forward modeling perspective, observing some physical phenomena 00:15:48.587 --> 00:15:51.907 in the brain and then modeling with ordinary or partial differential equations, 00:15:52.187 --> 00:15:59.287 seems to me that that approach in spite of being very successful in one domain has not generalized. 00:16:00.727 --> 00:16:04.087 It seems to me one needs a multiplicity of theoretical framework. 00:16:04.487 --> 00:16:10.287 So I have coined this motto, I say ontological monism and epistemological pluralism. 00:16:10.287 --> 00:16:12.627 So there is only one reality out there. 00:16:12.667 --> 00:16:15.047 We all agree. You always win at Scrabble, I understand. 00:16:17.927 --> 00:16:23.627 I don't think there's a disembodied soul that is driving my brain. 00:16:23.867 --> 00:16:30.187 But at the same time, I think there are multiple legitimate descriptions, 00:16:30.647 --> 00:16:33.287 epistemologies, ways of theorizing, 00:16:33.367 --> 00:16:35.767 modes of theorizing. They may use different mathematical tools. 00:16:36.287 --> 00:16:40.787 And it is not true that one reduces to the other. I think this is an assumption people have made. 00:16:40.967 --> 00:16:44.967 So when people talk about reductionism, the discussion is a bit confused because 00:16:44.967 --> 00:16:48.847 it's not whether they're talking about physical phenomena reducing to each other. 00:16:49.107 --> 00:16:51.847 That doesn't make any sense to me because there's only one phenomenon. 00:16:52.587 --> 00:16:54.807 The question is whether theories reduce to each other. 00:16:55.747 --> 00:16:59.847 And sometimes they do, sometimes they don't. And I think one has to be a little 00:16:59.847 --> 00:17:02.947 more Catholic. So you take a very Kuhnian approach here. 00:17:03.787 --> 00:17:07.227 I would like to, yeah, take a more Catholic approach to theorizing. 00:17:07.227 --> 00:17:09.167 And that is definitely new for me. 00:17:09.267 --> 00:17:13.887 I mean, I did not come from a tradition where that is done, although having 00:17:13.887 --> 00:17:17.327 trained as a condensed matter theorist, I was perhaps more open to it than if 00:17:17.327 --> 00:17:18.867 I were, let's say, trained in particle physics. 00:17:20.638 --> 00:17:26.198 Why do you think the mesoscopic scale is such an important scale to understand 00:17:26.198 --> 00:17:28.218 the brain at, as opposed to other scales? 00:17:28.338 --> 00:17:31.078 Or is it just that we haven't done enough work on that level? 00:17:31.078 --> 00:17:35.918 Yeah, I think that it is a scale at which the brain needs to be understood and 00:17:35.918 --> 00:17:38.658 a scale in which understanding is comparatively lacking. 00:17:40.198 --> 00:17:43.818 I'm also inspired to study this scale because it seems to be developmentally 00:17:43.818 --> 00:17:47.198 patterned. So it is in the genetic code, so to speak. 00:17:49.098 --> 00:17:54.178 Given that there is a juggernaut out there, which is kind of genome biology, 00:17:54.458 --> 00:17:59.678 and we have learned, certainly human knowledge has progressed in that domain. 00:17:59.678 --> 00:18:04.598 It would be great if we could link up neuroscience with that progress. 00:18:07.298 --> 00:18:12.938 One scale where I see that link occurring quite distinctly is at this mesoscale, 00:18:12.958 --> 00:18:15.458 because the mesoscale is developmentally patterned out of the genome, 00:18:15.578 --> 00:18:18.218 or at least that would be the quote-unquote hypothesis. 00:18:18.918 --> 00:18:21.778 So the heterogeneity at the mesoscopic scale, 00:18:22.018 --> 00:18:28.598 within elements at that scale, there might be Bohr-Humann homogeneity and the 00:18:28.598 --> 00:18:35.758 potential for using learning algorithms or developmental algorithms to wire up systems. 00:18:37.238 --> 00:18:42.338 Right, so the hypothesis would be that, let's say the genome patterns this mesoscale 00:18:42.338 --> 00:18:47.898 with environmental input, but let's say the genome, that's understood in modern 00:18:47.898 --> 00:18:48.718 day discussions, right? 00:18:48.898 --> 00:18:52.658 But without making those caveats, let us just say the genome patterns the mesoscale. 00:18:52.898 --> 00:18:58.198 Then variations in the mesoscale are, related to genomic variations, 00:18:58.438 --> 00:19:01.878 then one should be able to relate those two things, if one looks across species 00:19:01.878 --> 00:19:03.078 or if one looks across individuals. 00:19:04.173 --> 00:19:08.173 There could be also environmental perturbations that cause fluctuations at the 00:19:08.173 --> 00:19:12.133 mesoscale, but the idea is that it's genomically sculpted. 00:19:12.253 --> 00:19:19.553 And then variations at a more micro scale, certainly some of that patterning is also due to rules. 00:19:19.913 --> 00:19:24.033 Let's say one neuron has to connect to another neuron. So kind of self-organization, 00:19:24.133 --> 00:19:27.753 would you hope that that would give us leverage at the smaller scale? 00:19:27.953 --> 00:19:31.933 Right. One would expect that self-organization is playing a more important role at the smaller scale. 00:19:32.033 --> 00:19:36.113 So this would be a way of separating where the developmental program is more 00:19:36.113 --> 00:19:40.673 important from where the learning self-organization rules are more important. 00:19:41.033 --> 00:19:43.953 But what's the metric of this mesoscopic scale exactly? 00:19:44.653 --> 00:19:50.973 I don't think there is a fixed length scale that I would associate because if 00:19:50.973 --> 00:19:53.593 you look in different parts of the brain, there are different sizes. 00:19:54.173 --> 00:19:58.013 But I would simply point to classical neuroanatomical atlases, 00:19:58.353 --> 00:20:07.633 where the sizes of regions of nuclei vary if you look across the brain. 00:20:08.433 --> 00:20:11.833 They could get down to, depending on what brain you're looking at, 00:20:12.173 --> 00:20:15.353 in the mouse brain, they could get down to hundreds of microns, 00:20:15.533 --> 00:20:18.953 but they could also be larger, you know, millimeters. 00:20:18.953 --> 00:20:26.753 Meters so we have to we have to refine or improve this definition certainly 00:20:26.753 --> 00:20:33.093 if we want to do a comparative anatomy yeah i think it has to be data-driven i don't think there is a. 00:20:34.293 --> 00:20:37.173 Definition i can give today in advance of having 00:20:37.173 --> 00:20:41.453 gathered the data set what i hope is after we have data sets like this and we 00:20:41.453 --> 00:20:45.133 already have some existence proof from the gene expression data set from the 00:20:45.133 --> 00:20:52.433 allen institute that one can maybe empirically extract these correlation lengths 00:20:52.433 --> 00:20:54.833 or scales directly by looking at the data. 00:20:54.973 --> 00:21:00.693 What we can hypothesize based on neuroanatomical literature is that such scales exist. 00:21:02.319 --> 00:21:05.739 And, you know, good guesses as to what those scales are in specific brain regions. 00:21:06.099 --> 00:21:12.559 One thing you said earlier is that you felt that, let's say, 00:21:12.579 --> 00:21:17.639 the genome would be controlling development at just one of these scales. 00:21:17.779 --> 00:21:22.339 Well, let's say these lower levels would be more self-organizing. 00:21:22.859 --> 00:21:23.959 Why are you saying that? 00:21:25.319 --> 00:21:31.939 Well, it's a guess. Yes, I cannot articulate all my previous knowledge that 00:21:31.939 --> 00:21:37.119 leads to the guess, but one reason is you look at different mice, 00:21:37.419 --> 00:21:40.359 you get the same atlas. 00:21:40.559 --> 00:21:44.599 This is just a statement about how the community is organized in this research. 00:21:45.059 --> 00:21:49.739 If it was the case that there was tremendous environmental influence on the 00:21:49.739 --> 00:21:53.459 scale at which classical neuroanatomical atlases are organized, 00:21:53.819 --> 00:21:59.899 we would probably not be able to use atlases, So the very fact that we can use these atlases, to me, 00:22:00.039 --> 00:22:10.379 is some indication that these are not specifically environmentally sculpted. 00:22:10.379 --> 00:22:13.699 But this could also be a bias of how we treat these phenomena. 00:22:13.919 --> 00:22:17.579 Like in psychology, it's a typical problem that, okay, we move into statistics 00:22:17.579 --> 00:22:20.739 because we cannot understand humans from an individual level. 00:22:20.979 --> 00:22:25.219 On the other hand, now in psychology, people are bumping into this limitation 00:22:25.219 --> 00:22:29.579 statistical approach because actually there is not this normative person. 00:22:30.159 --> 00:22:33.779 So maybe for your emphasis, you're facing a similar kind of problem. 00:22:34.499 --> 00:22:39.879 I would certainly agree that there is not a normative mouse brain, the platonic brain. 00:22:40.379 --> 00:22:44.599 But what I do believe is that there's a low-dimensional manifold of brains, 00:22:44.719 --> 00:22:50.919 that it's not as high-dimensional a space as, let's say, the number of neurons 00:22:50.919 --> 00:22:52.539 in the brain would lead us to believe. 00:22:55.359 --> 00:23:01.239 So even though there is variability, and that variability could well be developmental 00:23:01.239 --> 00:23:05.939 plasticity that has environmental impact on it, in the same way that we would 00:23:05.939 --> 00:23:08.439 expect it in psychology. 00:23:10.739 --> 00:23:19.839 But I at least have a strong hypothesis that this pace of variation. 00:23:21.520 --> 00:23:27.840 Is quite constrained. It's quite small. And I would also say it differs from 00:23:27.840 --> 00:23:30.740 region to region in the brain, and we will probably find that out. 00:23:30.960 --> 00:23:34.800 This is an empirical question. We will be doing these studies in multiple mice, 00:23:35.000 --> 00:23:40.900 so we will be able to look across individuals and really ask the question empirically. 00:23:41.100 --> 00:23:42.000 I think that's the way to ask it. 00:23:42.300 --> 00:23:45.180 What's your feeling about this low-dimensional manifold? And certainly, 00:23:45.180 --> 00:23:49.600 if you look also across species or even including species that do not exist anymore. 00:23:50.020 --> 00:23:53.880 I mean, what's this common template there? What's your feeling about this? 00:23:54.180 --> 00:24:01.080 Are you expecting to converge to, say, a common design template for all vertebrates, for instance? 00:24:02.240 --> 00:24:05.960 It's a very intriguing thought, right? 00:24:06.040 --> 00:24:13.260 That there are, whether due to common ancestry or due to convergent evolution, 00:24:13.260 --> 00:24:15.380 solution, in some sense, 00:24:15.540 --> 00:24:21.540 selecting for the same circuit because the same function has to be subserved. 00:24:21.920 --> 00:24:27.740 It's a very intriguing hypothesis, which I hope is right, that there are these 00:24:27.740 --> 00:24:31.980 common design rules or templates that we will find. 00:24:32.680 --> 00:24:38.100 It may not turn out to be true, but that would be the interesting hypothesis in my mind. 00:24:38.240 --> 00:24:42.660 And it's an empirical hypothesis. this is, I would say that I hope it is true 00:24:42.660 --> 00:24:47.260 because there is certainly some theoretical biases I have that are pushing me 00:24:47.260 --> 00:24:49.580 in that direction. I think the system wouldn't work otherwise. 00:24:49.960 --> 00:24:52.480 Well, it gives them the finite set of genes building the brain. 00:24:54.271 --> 00:24:59.031 There you go. I mean, there is certainly some developmental rules, and I think that. 00:25:02.711 --> 00:25:08.771 We have other examples like our musculoskeletal system that are also patterned 00:25:08.771 --> 00:25:10.471 about which we don't ask these questions. 00:25:10.991 --> 00:25:14.931 So is the brain really that different from all the other organs in the body 00:25:14.931 --> 00:25:19.311 at the level, let's say, of this mesocircuit? 00:25:20.731 --> 00:25:23.751 So your idea is to do a lot of the work in 00:25:23.751 --> 00:25:26.911 your own laboratory at Cold Harbor but you're 00:25:26.911 --> 00:25:29.831 trying to get other people on board this project because 00:25:29.831 --> 00:25:35.911 by the sounds of it it expands to include many other species also to include 00:25:35.911 --> 00:25:41.651 animals at different ages there's a huge work program here why do you think 00:25:41.651 --> 00:25:45.411 people should prioritize this as opposed to all the other things that we might 00:25:45.411 --> 00:25:49.451 be doing in the neuroscience well I guess you you know, it's a free world. 00:25:50.691 --> 00:25:56.051 I have to choose from that side. No, no, no. 00:25:56.091 --> 00:26:02.411 But more seriously, I think that, I don't know, talking about humility, 00:26:02.671 --> 00:26:04.991 I've just come to realize that the problems are extremely hard. 00:26:05.131 --> 00:26:10.531 So I would like to work on areas where clearly progress can be made. 00:26:10.611 --> 00:26:11.671 Now that can be very dangerous. 00:26:11.871 --> 00:26:14.371 One can get into instrumental fallacy, right? Right. 00:26:15.231 --> 00:26:22.231 But I have tried to articulate some of the arguments in the position paper that 00:26:22.231 --> 00:26:24.651 we put together that I would draw people's attention to, 00:26:24.711 --> 00:26:31.891 plus computational biology that came out last year with the first author, 00:26:32.071 --> 00:26:34.191 Jay Boland, and the last author being myself. 00:26:34.191 --> 00:26:37.931 Self um but just to 00:26:37.931 --> 00:26:41.591 you know try to lay out some motivations 00:26:41.591 --> 00:26:44.411 uh one great 00:26:44.411 --> 00:26:47.531 motivation is that we are at the technological cost where 00:26:47.531 --> 00:26:52.971 this has just become doable within the funding constraints of individual laboratory 00:26:52.971 --> 00:26:58.071 i think that's for me a great motivation and that was not true a few years ago 00:26:58.071 --> 00:27:02.591 simply due to memory costs today it is doable yeah but this is the because we 00:27:02.591 --> 00:27:04.931 can't argument so Because we can't argue with that. 00:27:04.971 --> 00:27:06.191 I think it's an important argument 00:27:06.191 --> 00:27:08.551 because there are many… You can win elections with that, you know. 00:27:11.871 --> 00:27:16.651 That's it. No, I mean, I think it's an important argument. 00:27:17.711 --> 00:27:21.811 But the second argument, I think, is that…. 00:27:24.609 --> 00:27:26.289 The scope of discovery, I think, 00:27:26.289 --> 00:27:37.809 is large. We simply don't have these circuits at a very simple level. 00:27:37.889 --> 00:27:41.169 How are we going to understand how the brain works? The brain is a circuit. 00:27:41.589 --> 00:27:45.089 We don't have the circuit. People could argue we know certain bits and pieces 00:27:45.089 --> 00:27:47.449 of it, but that's a very simple argument. 00:27:48.669 --> 00:27:51.229 One can try to refine the argument, but that's one argument. 00:27:51.229 --> 00:27:55.129 And then there's a third argument, which is this information integration that 00:27:55.129 --> 00:28:00.709 we were talking about, that we are battling this huge deluge of data and information. 00:28:00.849 --> 00:28:03.789 How to tie it together? Well, here's a way we could. 00:28:05.269 --> 00:28:11.909 And it's clearly a project that is much bigger than something that I could handle in my laboratory. 00:28:12.169 --> 00:28:14.889 So I'm definitely hoping that other people will want to join in. 00:28:14.889 --> 00:28:19.409 And we are very open, by the way, with our data set, even before publication, 00:28:19.629 --> 00:28:24.209 when the data will officially be made public, if people would like to collaborate 00:28:24.209 --> 00:28:28.149 with us at the level of data analysis, we're happy to do that. 00:28:28.189 --> 00:28:31.609 If people would like to learn how to do these experiments, or if they're interested 00:28:31.609 --> 00:28:36.569 in setting up their own pipelines, we're also very happy to actually provide 00:28:36.569 --> 00:28:37.569 that input and expertise. 00:28:38.009 --> 00:28:41.569 Because I think that the reality is that we'll publish a two, 00:28:41.689 --> 00:28:45.429 three-page paper, let us say, but that won't summarize what's going on. 00:28:45.669 --> 00:28:49.449 Things are changing in that way. However, one thing I was wondering about now 00:28:49.449 --> 00:28:55.349 is whether you're not stepping a bit too easily over this challenge that also Tony put in front of you. 00:28:55.409 --> 00:29:01.849 Imagine now we have this whole federation of laboratories contributing to this 00:29:01.849 --> 00:29:04.429 pipeline, and they all have their own species to work on or whatever. 00:29:06.046 --> 00:29:09.406 You have a huge problem integrating all this data because at some point people 00:29:09.406 --> 00:29:13.426 would like to bring in other kinds of data, physiological data, behavioral data. 00:29:13.666 --> 00:29:20.306 Do you really have a taxonomy in place to bring all that information together in an integrated form? 00:29:20.566 --> 00:29:25.106 Or do you think that's just going to self-organize? It's a work in progress. 00:29:25.846 --> 00:29:28.366 Some people are working on it. 00:29:29.366 --> 00:29:32.526 I do think it has to be, to some extent, data-driven. 00:29:32.686 --> 00:29:38.466 And talking about theory, here is one place I think theorists really need to pitch in. 00:29:39.286 --> 00:29:43.406 Because taxonomy is really a theory about the system. 00:29:43.526 --> 00:29:47.766 Once you start dividing what the parts are and naming them, you've kind of made 00:29:47.766 --> 00:29:50.726 your mind up about how to divide the system and how to name them. 00:29:51.146 --> 00:29:54.806 But isn't it a scary thought to actually now open the floodgates, 00:29:54.946 --> 00:29:57.506 but you don't have the buckets ready yet to catch the water? 00:29:57.506 --> 00:30:04.266 I think that there are some starting efforts. 00:30:05.126 --> 00:30:10.926 So it's not as bad as it could be in other fields. Because in anatomy, 00:30:11.066 --> 00:30:12.226 there has been a scholarly tradition. 00:30:13.606 --> 00:30:18.486 And so there is certainly a good head start, I would say. 00:30:19.586 --> 00:30:23.206 But you're right. I mean, it is a challenge. 00:30:24.826 --> 00:30:31.166 I think that we will be helped simply by having unified datasets. 00:30:31.466 --> 00:30:38.546 This taxonomy problem, what we want to call a region will be less important 00:30:38.546 --> 00:30:44.086 than having a coordinate where you can really say that, well, I call it A, you call it B, 00:30:44.286 --> 00:30:46.646 but we are all referring to the same objective, 00:30:47.925 --> 00:30:50.865 artifact and that's what will help that taxonomy problem 00:30:50.865 --> 00:30:53.965 so what was amazing in the human genome project is 00:30:53.965 --> 00:30:57.205 that there was this expectation that would take quite some 00:30:57.205 --> 00:31:02.025 time to sequence human genome and then actually some people really industrialize 00:31:02.025 --> 00:31:05.665 this whole process and it turned out could be done just in a few years which 00:31:05.665 --> 00:31:09.365 was astonishing even though it's still debatable how informative all this data 00:31:09.365 --> 00:31:14.925 is now how many many years is it going to take you to to get the data for, 00:31:15.225 --> 00:31:18.025 let's say, a single mouse species? 00:31:18.625 --> 00:31:20.305 How many men years are we talking about? 00:31:21.125 --> 00:31:24.305 Well, we are hoping to release a draft next fall. 00:31:24.885 --> 00:31:29.225 And there's about... Now, what I mean by... But I meant complete, right? Everything. 00:31:31.325 --> 00:31:36.885 Everything will take infinitely long, by definition. The whole brain. 00:31:37.045 --> 00:31:42.265 But I think on the scale of two to three years, 00:31:43.305 --> 00:31:53.725 we will have given existing projects we will have a quote unquote mesocircuit even next year we will, 00:31:54.605 --> 00:32:01.185 you know in every laboratory order 10 people maybe so you have 30 man years 00:32:01.185 --> 00:32:05.765 you said 30 man years to get one mesocircuit of one species done. 00:32:05.925 --> 00:32:11.185 I think that's reasonable but we have defined mesocircuit in a particular way 00:32:11.185 --> 00:32:13.825 People are going to challenge that and have arguments about it. 00:32:15.845 --> 00:32:18.665 Yeah, I think it's a doable project on a short time scale. 00:32:19.065 --> 00:32:22.325 And for yourself, I mean, is there going to come a point where you say, 00:32:22.405 --> 00:32:26.245 okay, I've got enough data of a sufficient quality. 00:32:26.965 --> 00:32:30.385 I'll go back and be a theorist about this data. 00:32:31.985 --> 00:32:37.505 It's happening already. Okay. You know, in the sense that even when trying to 00:32:37.505 --> 00:32:41.085 interpret our very first data sets or in trying to analyze them, 00:32:41.185 --> 00:32:44.645 the problems that come up are definitely of a conceptual theoretical nature. 00:32:44.805 --> 00:32:45.625 I'll give you an example. 00:32:47.925 --> 00:32:51.365 This question came up about quantifying connection strength. 00:32:52.565 --> 00:32:58.225 You often see isolated neurons in different parts of the brain showing up. 00:33:00.285 --> 00:33:05.445 If you have hundreds of thousands or tens of thousands of neurons in a particular, 00:33:05.445 --> 00:33:09.885 or let's say region of the brain that you have injected, and if one neuron shows 00:33:09.885 --> 00:33:11.965 up in some other part of the brain, 00:33:12.745 --> 00:33:18.185 in some sense, the circuit is, or doesn't have this clean organization that 00:33:18.185 --> 00:33:19.505 we theoretically started out. 00:33:20.891 --> 00:33:22.091 That it has got these tails. 00:33:24.331 --> 00:33:27.351 How do you even think theoretically about a circuit like that? 00:33:27.391 --> 00:33:29.471 What are those individual neurons doing? 00:33:29.591 --> 00:33:33.791 Are they artifacts of the fact that biology is not a clean engineer? 00:33:34.611 --> 00:33:37.031 Or was it a reason? Is it a feature or a bug? 00:33:38.191 --> 00:33:44.571 But now in this endeavor, which also your project got supported to go after 00:33:44.571 --> 00:33:45.791 this connecto of the brain, 00:33:46.491 --> 00:33:50.191 there are also other initiatives underway where people would look more at, 00:33:50.191 --> 00:33:54.471 Let's say what we call the effective connectivity of the brain based on data 00:33:54.471 --> 00:33:59.771 sets that are of a very different nature, using different kinds of imaging techniques and so on. 00:33:59.971 --> 00:34:03.291 Do you expect a convergence among these different approaches, 00:34:03.451 --> 00:34:06.051 or do you think that's just too far in the future? 00:34:06.391 --> 00:34:10.131 I have a theoretical disagreement with, and I'll be very frank about this, 00:34:10.271 --> 00:34:17.331 about the term functional connectivity being used to describe what are temporal correlations. 00:34:17.331 --> 00:34:24.071 So, simply because two time series are correlated, and this we have known for 00:34:24.071 --> 00:34:29.471 a long time, does not mean that there is a, let's say, anatomical connection 00:34:29.471 --> 00:34:32.171 between those two points. 00:34:33.491 --> 00:34:37.491 So, I don't see any reason why there should be a quote-unquote convergence. 00:34:37.811 --> 00:34:43.371 It's certainly true that the anatomical connectivity is going to drive correlations, 00:34:43.691 --> 00:34:46.631 but I don't see the reverse rub. 00:34:47.331 --> 00:34:50.631 So the correlations are a necessary condition of the causal structure, right? 00:34:50.911 --> 00:34:56.791 Of the anatomical structure. And I strongly advocate calling correlations correlations. 00:34:56.931 --> 00:35:03.811 I think it is misleading and perhaps even distracting to call it connectivity. 00:35:04.591 --> 00:35:07.091 Your database will be monosynaptic. 00:35:08.217 --> 00:35:12.077 Our database is really of neuronal morphology. We don't really have synapse. 00:35:12.257 --> 00:35:16.237 Oh, okay. But you'll be just looking along a single fiber. 00:35:17.817 --> 00:35:22.517 So it'll still be a problem for me if I want to know if there's a connection 00:35:22.517 --> 00:35:25.397 between A and B that might go via C. 00:35:26.617 --> 00:35:31.097 I can tell there's a projection that goes into C, a projection from C into B, 00:35:31.197 --> 00:35:35.257 but I can't know from your database that those two things meet up. Correct. 00:35:35.537 --> 00:35:39.977 And there are other limitations to this project. Namely, we are not being cell-type specific. 00:35:40.597 --> 00:35:44.797 There are other projects which are coming about which will try to be more cell-type 00:35:44.797 --> 00:35:48.217 specific and yet get neuro-anthropomical connectivity in place. 00:35:49.117 --> 00:35:56.317 I regard this particular mesocircuit that we are mapping out as operationally defined. 00:35:56.597 --> 00:36:00.237 Operationally because it's defined in terms of retrograde and anterograde injections 00:36:00.237 --> 00:36:06.417 and whole-brain imaging object. object, but there will be missing pieces of 00:36:06.417 --> 00:36:11.037 information that will then have to be layered onto this thing, more projects. 00:36:12.197 --> 00:36:16.957 Now, also in your presentation, also in your work, you have been spending quite 00:36:16.957 --> 00:36:19.577 some time to look at gene expression patterns in the brain. Yeah. 00:36:19.977 --> 00:36:25.857 And do you see that as giving you leverage in understanding these anatomical 00:36:25.857 --> 00:36:30.477 pathways that you're trying to reveal right now, or do you see this really as 00:36:30.477 --> 00:36:32.457 separate databases as well? 00:36:33.017 --> 00:36:38.317 No, I see the connection that they are both indicating a meso-structure, so to speak. 00:36:38.517 --> 00:36:41.917 You know, they are getting us objectively at this intermediate length scale 00:36:41.917 --> 00:36:43.657 that is there in neuroanatomy. 00:36:43.997 --> 00:36:47.897 The gene expression data sets are probably giving us indirectly information about cell types. 00:36:48.517 --> 00:36:50.637 I don't think they are directly telling us about connections. 00:36:52.481 --> 00:36:57.441 But yes, they are definitely connected. So within this gene expression data 00:36:57.441 --> 00:37:00.221 you showed, there was something very puzzling because anatomically, 00:37:00.481 --> 00:37:03.381 at least if I listen to an anatomist, I'm not an anatomist myself, 00:37:03.581 --> 00:37:06.501 they would always tell you, look, cortex is relatively uniform, 00:37:07.041 --> 00:37:09.141 well-structured, modular, etc. 00:37:09.881 --> 00:37:15.181 And if you go to subcortical areas, things get more messy. It's more variable, 00:37:15.481 --> 00:37:17.081 heterogeneous, and so on. 00:37:17.961 --> 00:37:21.321 And if I look at your gene expression data, you made the point. 00:37:21.861 --> 00:37:25.521 Actually, you have less genes expressed at the subcortical levels than at the cortical levels. 00:37:25.921 --> 00:37:29.881 So what do you make of this apparent contradiction? 00:37:30.621 --> 00:37:33.841 I actually think it's the other way around. I think the subcortical structures 00:37:33.841 --> 00:37:38.921 are more homogeneously structured and more carefully sculpted. 00:37:39.701 --> 00:37:44.101 You have much more interspersion of cell populations, also the kind of transmitter 00:37:44.101 --> 00:37:47.221 systems. Correct. So they are more... Yeah, I see what you're saying. 00:37:48.021 --> 00:37:52.181 It's more messy in the sense that the modules are not well separated, 00:37:52.361 --> 00:37:54.821 but it may be less variable. 00:37:58.321 --> 00:38:03.061 So, yes, it is messy. It will be messy for us to disentangle, 00:38:03.241 --> 00:38:07.421 but it will be less variable, I think, from animal to animal. 00:38:07.721 --> 00:38:11.901 So sometimes what you're saying is maybe cortex is so neatly organized because 00:38:11.901 --> 00:38:16.001 there are just more genes being expressed there to keep that clean structure in place. 00:38:16.761 --> 00:38:20.401 While you would have less genes expressed subcortically, leading to more of 00:38:20.401 --> 00:38:24.461 a dispersion of how these cells actually anchor themselves in the substrate and so on. 00:38:24.521 --> 00:38:27.361 Is that how I interpret what you're saying? That's an interesting hypothesis. 00:38:29.721 --> 00:38:35.161 I guess the cortex is not translationally homogeneous, as we know from Brodmann. 00:38:35.421 --> 00:38:39.501 Areas in the gene expression pattern certainly show that there are different... 00:38:41.129 --> 00:38:46.629 Vertical structures. Cool. So look, I have two questions to finish up. 00:38:48.129 --> 00:38:49.649 If we had to, 00:38:51.809 --> 00:38:58.509 define the Partha law of science, our attempts on the brain, 00:38:58.789 --> 00:39:02.809 what would be this Partha-Mitra law of investigation and understanding? 00:39:03.069 --> 00:39:04.609 What's the law we should adhere to? 00:39:05.109 --> 00:39:08.469 Law? Well, I made my aphorism, right? 00:39:08.969 --> 00:39:14.069 There's one reality about multiple theoretical approaches, I think we have to 00:39:14.069 --> 00:39:19.169 be open-minded about disciplines. 00:39:19.469 --> 00:39:27.689 And this sounds like a truism, and it sounds really something that people say as a slogan, 00:39:27.849 --> 00:39:32.689 but I really think that we need to educate ourselves in different disciplines 00:39:32.689 --> 00:39:36.829 like engineering disciplines or neuroscientists need to learn more about population 00:39:36.829 --> 00:39:38.669 biology and evolution and so on and so forth. 00:39:38.669 --> 00:39:42.149 Perfect and then so so if i 00:39:42.149 --> 00:39:45.229 go visit you again in cold spring harbor five 00:39:45.229 --> 00:39:48.229 years from now i'm gonna say okay there's this 00:39:48.229 --> 00:39:50.789 one prediction you gave to me today and today i'm going 00:39:50.789 --> 00:39:53.529 to check whether it's true or not what's this one 00:39:53.529 --> 00:39:57.129 prediction you're willing to stick your neck out today well that's 00:39:57.129 --> 00:40:03.249 a very tough one come on you're a tough guy um i 00:40:03.249 --> 00:40:10.029 guess this notion that the meso circuit is genetically wired and that genetic 00:40:10.029 --> 00:40:16.449 polymorphisms will cause alterations in the mesocircuitry that may be correlated 00:40:16.449 --> 00:40:17.989 with neuropsychiatric disorders, 00:40:18.329 --> 00:40:21.369 which also have genetic predispositions. 00:40:21.669 --> 00:40:27.229 That's for me, especially the affective systems that are related to emotional behaviors. 00:40:27.749 --> 00:40:33.169 That for me is sort of the, I don't know if it's a prediction, 00:40:33.329 --> 00:40:41.489 but that's a story that I hope in five years to really understand and gain insight into our behaviors. 00:40:41.849 --> 00:40:44.789 Wonderful. Parthamitra, thank you very much. Thank you.