A highlight from Prof. John Pearson, Assistant Professor of Neurobiology at Duke University
My guest today is professor john. Person who's assistant professor of neurobiology. A duke today is research focuses on application machine. Learning methods do the analysis of data malcolm. Hi thanks doing the so I want to start with one of your recent papers. online estimation but noisy testing. You say one of the primary goals of systems euroscience is to relate the structure of your surrogates to their function yet. Packets of connectivity are difficult establish than according from large populations indicating organisms. Can please approaches. Have attempted to estimate fox connectivity between neurons using statistical modeling of original data. But you say about this approaches. Like heavily on parametric assumptions and a purely relational so so the noodle connectivity obviously if the galvastan that well has office applications system Before we get into the details. John hong have things crudest sort of look backward lost in fifteen twenty years. Are we making progress in this area. I think it's interesting. It's one of those questions that it would depend a lot on who you ask. So maybe one of the places to start is a really classic set of studies by astro princeton eve marder and they took this somatic gastric gangling to really simple circuit and only has a few neurons and this was now not quick. Twenty years ago they asked and they said well we know all the connections in the circuit and what are the different ways that neural activity in the circuit can behave and with found out as it was an extremely complicated question. And so i think one of the things that neuroscientists have had to keep in mind is that structure and function are not the same so what they found in this in this tiny bundle neurons from the crab is that Depending on temperature or depending on how things were stimulated the patterns of activity in that little tiny circuit were very very different and you can qualitatively weren't similar to each other and so on the one hand we have this counter example. That says even if you tell me the structure of his circuit. It's not clear by that alone. What function is running into the same thing in that we've had to. I would say really. Big advances in the last few years. One is amazing. Data set of human brain connectivity data. And there. I don't mean that we actually know anatomical connectivity but what you would get from in our mri scanner so diffusion tensor imaging which gives you some measures of connectivity so really large samples in humans with the human connect project. The other is a really fascinating project is only partway through and this fly. Em project from genelius Where they are going to trace all the connections in the library and that has been a massive undertaking and they released a piece of it. I think last year and they are readying the rest of it for the for the next couple of years. And there's a both in big advances. But i think there's a very active debate in the field about more. We're going to use the data four at and that's i think where we come to part of where the genesis for this study was is. We're still sort of stuck at neuroscience with in one or two cases all that connectivity data we get is from saint many many flies all of which are dead and the new recording data we get is from living organisms but we don't know exactly how things are wired up and not being able to get both of those things i think has a lot of groups not just hours interested in ways that you can do causal experiments in the circuit you can poker prod something nippy late the activity in the circuit and learn something about how the structure and function relationships are tied together. Yes oh it's annoying. Boggled john in a he's interesting do to think about sort of the relationship between as you as torture and function. I would think that the structure could be sort of random mutations the right if other processes might be happening and then ultimately you make use off the structure your given to a function over specialization. Ease that debate to think about. The brain is structurally walking from the function or structure. Sort of give me to function over time. Well it's it's the chicken and egg problem and of the reason it's that way he is on the one hand we know that these circuits are extremely plastic so going back to extremely classic work by donald hebb if two neurons co activate each other connections get stronger and we know that that's a substrate for all sorts of learning that happens in the brain but the other thing that i would say we know. Is that the process of development. Very exquisitely patterns certain circuits so there seemed to be levels. Maybe maybe the level of what people in the artificial neural networks would call architecture. Where the connectivity patterns really you need to get them right and maybe you don't need to get them right in terms of exactly which is connected to exactly what other but maybe which types of neurons get connected up and what sorts of motifs uc circuits. That's incredibly important. And so all sorts of genetic factors during development.