 Thank you, Sean, and the other organizers for the invitation and to the excellent INCF staff who's made this such a wonderful conference. And I'm especially happy to be talking in this mini symposium because I think it's almost certainly the first symposium that brings together the Allen Institute that's been in the business of large-scale neuroscience and neuroinformatics projects for the past decade and is just at the beginning of its second decade expansion into into new fields that I'll be telling you about. But it's wonderful to be along here with the human brain project from the EU and the United States nascent brain initiative. So I'll be telling you about work that's just beginning at the Allen Institute, a project called Project Minescope, exploring mammalian neoportex in a high throughput manner. And I think it's useful to go back to the beginning and the present at the Allen Institute. Again, it's 10 years old. It's having a 10 year, 10th anniversary celebration this month or next month, September. And it's really important to emphasize the great contributions and accomplishments of the Allen Institute over the past decade starting with the mouse brain atlas. This is the portal. I suppose I don't have a, there's the pointer. This is the Allen Institute's portal BrainMap.org. And the first product, thank you, the first product in BrainMap.org is the mouse brain atlas which is a map of the gene expression in the mouse that was completed over the first four or five years of the Institute's existence which has been extended greatly in the ensuing five or six years to have a gene expression atlas developing mouse brain of non-human primates of the human brain and atlases of development. And all of these have really been great advances in neuroscience that have been of great use to the neuroscience community but also really signal achievements in neuroinformatics in large databases that are online that are easily queryable by the public and the neuroscience community. Most recently is the connection atlas, an atlas of the connections between different brain areas in the mouse that I'll be telling you about briefly. But my primary task today is to tell you about the Minescope project which is a large component of what the Institute is doing currently and will be doing over the next decade but it's only one component. And the Minescope program is a program to examine the mouse visual system with a concentration on the cortex and associated thalamic nuclei and really going from photons in from visual stimuli through behavior. And I think I'll read the mission statement. We seek to understand the computations that lead from photons to behavior by observing and modeling the transformations of signals in the corticothalamic visual system. We want to catalog and characterize the cellular building blocks of the cerebral cortex, their dynamics, and cell type specific structural and functional connectomes. We want to know what the animal sees, how it thinks, and how it decides. So there's a real emphasis on explaining in vivo physiology all the way through behavior. And as I'll tell you it has three components. Neural coding, that's the group that I'm meeting, cell types led by Hong Kui Zhang who's been at the Institute for around six years, and modeling analysis and theory led by Christof Koker, a chief scientific officer, and Mike Harleys who's been at the Institute almost from the beginning. So the Minescope team really is a it's a team. Starting all the way at the top with Alan Jones our CEO, Christof Koker, chief scientific officer. I've told you about the three of us leading the three equal-thirds of Minescope. But in addition, Qin Dang who leads technology, John Phillips who is not listed, who leads structured science and large-scale projects, and a large number of PhD level scientists starting with investigators, junior group leaders including people in all three divisions of Minescope. And within a year we're going to have 52 or more PhDs working in this project. That'll be within roughly two years of its inception. And all of this was made possible by really an unprecedented gift from Paul and Jodie Allen who have made a significant pledge for the first four years of this project after a decade of support of the Allen Institute. And really that's the inception of what promises to be a 10-year project. So the goal of Minescope is to understand coding and what do we mean by that? How did the 1 to 2 million nerve cells in the mouse visual cortex, not in primary visual cortex, but at all in the visual cortical areas represent and transform visual information? Starting with describing the building blocks, the cellular components, the cell types, and their projections. Counting, just enumerating the components with quantitative neuroanatomy. Recording, this is in the neural coding component. Observing the spiking activity of neurons of each of these cell types. Interfering using optogenetic tools, using light to modulate the firing of neurons to see how this affects the firing, the physiology of the circuit, and all the way through behavior. And then finally, and very importantly, bringing it all together with modeling, predicting and modeling these neural networks in a quantitative manner. And theory, this movie was supposed to play earlier. This is a movie showing the physiology, and I'll explain it later. I'll stop that now. And understanding, really, it's not just merely a data generation exercise. It's really a group endeavor trying to achieve understanding and theoretical understanding of the mouse visual system. And really, more generally, mouse cortical circuits, both single circuits, primary visual cortex, and the network of networks of the multiple different visual cortical areas. And what's unique, I think, in the organization of this is that it's injected into the highly successful organization of the Allen Institute. And it's really modeled after biotech as a matrix organization, that there are projects that draw from the resources at the institute. So the project is Project Mindscope, which again has three components, neural coding or experimental neural coding, cell types, and modeling analysis in theory. And these groups are decent, quite big, with primarily PhD and just a few technicians, but most of the experimental help and computational help comes from central resources that are organized separately and shared through the institute. So there's a very important program management component, Structured Science, which is the group that has already brought all of these wonderful projects to fruition over the past decade. Technology, which is the information technology, computational engine that runs all of the analysis, all the way through the web products, and of course, engineering, allowing the experiments to be performed. And we decided very early, sort of the ground rules of this endeavor were that it obviously should be in the mouse with all of the wonderful genetic tools that are available in mouse and the great strengths in mouse neuroscience at the institute, and that it would be cortical centric, that it would really attack the problem of cortical computation and that it would relate to behavior. And quite early, we settled on the mouse visual system, which really is a wonderful system that, although the mouse has low visual acuity, it has actually a quite wonderfully developed visual system. And it's very similar to the visual system of other mammals, but writ small, with large but more tractable numbers than other species. 45,000 retinal ganglion cells, which are the output of the retina. 18,000 LGN neurons at the LGN at the lateral geniculate nucleus of the thalamus, the primary relay station for information from the ganglion cells to primary visual cortex or V1. And as opposed to primates with probably hundreds of millions of cells in primary visual cortex, or many millions of cells in primary visual cortex, the mouse primary visual cortex has only 75,000 or 100,000 neurons in layer 4, that is, in layer 4, and 360,000 neurons overall. The reason I concentrate on layer 4 in this slide is that our first one or two-year milestone project to show that the whole system is working is an in-depth analysis of the primary input layer of cortex layer 4. I'm only barely going to touch on that. So really for the purposes of this talk, I'll be discussing primary visual cortex and then the network of higher visual areas, which there are 14 million cortical neurons in the mouse, and one or two million of those are related to the visual system. And it's really a wonderful part of mouse neurobiology that over the past decade, led by Burkhalter at Washington University, a detailed map of the mouse visual cortical system, where there are 10, maybe 12 distinct areas that process visual information, all of which are arranged in a hierarchical manner, much like the hierarchical visual system that one sees in the primate. Again, writ small, and the details are different, but it really provides a wonderful model system of a network of networks that allows us to study cortical communication. So the dominant metaphor that Christoph Koken and I and the rest of us at the Institute like to use is that of astronomical observatories. And Christoph and I wrote a piece in commentary in Nature, I guess, two years ago, where it was entitled Observatories of the Mind. And using the metaphor of astronomical observatories, just looking at this next generation telescope, a ground-based telescope with 30-meter primary mirror consisting of many hexagonal mirrors. This is a very large-scale project that will bring astronomy to the next generation 10 times higher resolution than Hubble. It's a very long-term project. It's ballpark similarly expensive as all three of the endeavors that we'll be talking about today. And we want to do something similar. We want to create observatories of the mind that have a number of attributes so that one would usually expect with big science. Standardizable, reproducible, of course, accurate. We'll be putting data online. The first three sort of standard reason that's the way science should be, but I think neuroscience, everyone included. We do small-scale studies, and it's often very difficult for labs to reproduce the work sometimes of themselves, but certainly of other laboratories. And it's wonderful to have a framework and an institution where that kind of reproducible science is what they do. And along with that, it is scalable that in individual laboratories, one can always do a small number of experiments leading to a paper. But that doesn't mean that those techniques can be scalable, which means teaching the experimental techniques to a significant number of people so it can be run the same experiment day in and day out. And I'll show a recent example of that from the Allen Institute. And that's the connectivity atlas. So the connectivity atlas is a project that's led by Han Kui-Zang, who also leads the mouse cell types component of the Minds Code project. And it's being performed by the structured science team at the Allen Institute. And this atlas is, again, a single experiment that has been performed by numerous laboratories by using viruses to trace the projections, the connections from one area to the next. And either using a virus to infect all local cells where it's injected and see where those cells project, or to infect only certain cells that are driven by Cree, that express Cree. And an enzyme that allows another transgene to be expressed. And these Cree lines can be many different things, but the goal of this part of neuroscience is to have lines that are as close as possible to a well-defined neuronal cell type, such as pyramidal cells in layer 5 that project the superior calliculus, for instance. And the later stage of this connection atlas is using cell-type specific Cree driver lines. So a virus is injected into mice of a specific age, all standardizable. It's put into a machine that does serial sections through the brain every 100 microns. And the brain itself is imaged with two-photon microscopy. And this is a machine that allows an entire mouse brain to be reconstructed at this resolution in less than 24 hours, using almost a terabyte of data from each mouse. And the important thing is that with standardized injections, very standardized brain placement and registration techniques, all neuroinformatic techniques, currently it has more than 1.2 petabytes of anatomical data online that is viewable with a standalone viewer 3D brain explorer, or it can be looked at through the web portal, as I showed you. And this is a four-year project that's coming to completion and the next generation is just beginning now. And it's online already, one of the initially scary, but as one joins the Institute, but really wonderful thing is that for all of these projects, the data don't go online after the paper is published. The data go online as soon, not as they're taken, but as soon as they're validated and go through quality control, registration, et cetera, they go straight online. And so before the publication, there are hundreds of neuroscience throughout the world who regularly go to this resource to look at their area of their brain and see what the projections might be. So here's an example of a number of injection sites in the dark, the bright colors of the injection sites and then it's hard to see here, but you know, go to brainmap.org, I implore you, it's really a wonderful thing to wander through the data and you can look at all the projections from the various areas, but there are these wonderful informatics tools that allow you to visualize the injections and the projection sites in three dimensions within the Allen brain, mouse brain coordinate system. And from these data, you know, it's standardizable, reproducible, scalable, but it's very important, it's complete in some sense. Of course, you can't ever be totally complete, but there's a very rich and well thought out targeting strategy so that at the end of the day, there is a connectivity matrix from hundreds of distinct brain areas, from hundreds of areas to hundreds of areas and really this object is unprecedented, I think in neuroscience, a really detailed connection atlas throughout the mouse brain with quantitative data and the amplitude, the strength of those connections measured in the fluorescence of the terminal fields scans, I forget but it is, but certainly three or four, maybe five orders of magnitude. And it can be examined by theoretical neuroscientists to look at the structure of such networks, model dynamics on such networks, but it really also can be used by classical neuroanatomists or people with specific questions. For instance, people interested in thalamic to cortical interconnections. So that's the precursor of one part of the cell types program. The rest of the cell types program includes many components and I can just list them here. I really won't have time to go through all of them, characterizing the cell types in V1, imaging and reconstructing morphology of single neurons in V1, map the connections using both anterograde and transnaptic tracing, electrophysiological profiling, which I won't be able to tell you about and transcriptomic profiling, which again I won't be able to tell you about but here's some transcriptomic data. I'm going to concentrate in the next 10, 15 minutes on the neural coding program just to give a deep dive into one program that, you know, the program that I'm most familiar with, you know, with the program that I'm leaving and I'll just give an overview of the motivation of the program and two types of data that we will be collecting. But really I want to step back and motivate studying the visual system in any species and just for people who haven't seen sort of the basic properties or more importantly listen to the basic properties of visual portable neurons, I think it's always best to go back to the source and it's very fitting the Swedish source here in Stockholm, David Hubel and Torsten Wiesel, their work starting in the late 50s. And but for the purposes of neural coding, we really want to understand two things, you know, what individual neurons do, you know, what is the computation and the much more difficult question, how do they do it, you know, what physical structure is used to perform these computations. But starting with, you know, what do they do, the neurons do many things but among the things they do is respond to visual stimuli and the quintessential new feature in visual portable responses compared to most of the retinal and phylamic afferents is orientation selectivity. So I just want to show this wonderful movie taken in the 70s by Hubel and Wiesel. And can everyone hear that? So what you're hearing are the action potentials of a single neuron in the visual cortex. And you can hear that it fires action potentials to a visual stimulus that is in quote the right place and now they're drawing a box which is the receptive field, the area in visual space that if the stimulus is there spikes maybe a vote. Can I be turned up? Electrophysiology is always hard. Now I've gone back to the beginning of the movie. Again, we're going to the beginning of the movie. This is where the, I think this was working. There we go. Just in time. So a stimulus in the right place evokes spikes but it's a very selective receptive field that many spikes are evoked if the stimulus is of the right orientation. But here comes the critical experiments. Few spikes and now no spikes at all. And now we have spikes again. So that's only one of the many computations that a cortical network might perform. And you know, that's the view in the 60s but now there's a much more sophisticated, much richer view of receptive fields. But the story stays the same. One is the story is that given the responses of neurons and also the behavioral modulation of those responses how can we explain those responses? And now we have more advanced tools such as calcium imaging. This is a movie from years ago from work that we did at Harvard where we're looking at 100 neurons in the cat visual cortex and the neurons are the bright spots and when they're activated they get yet brighter. And you can see that when the stimulus moves down to the left half of the neurons fire action potentials and up into the right the other half. So this is showing a map of direction cell activity in the visual cortex. And the goal of the neural coding program is to, through a series of physiological and anatomical observatories study the relationship between structure, connectivity, cell types and function in the mouse visual system. And going back to the last sentence in the mission statement of Minescope we want to see, to understand how the animal sees, how it thinks and how it decides all the way through behavior and decision making. And again this is through a number of observatories. And I like going through this series of images of the mouse at successively finer scales to show how different observatories might look at the visual system at different scales. We start off with the entire mouse, obviously an intact mouse that's actually awake and behaving or reacting to visual stimuli and we look at its brain or we listen in with electrodes I won't be telling you about. We look at the brain through a small portal in the brain, a glass window that allows us to look directly at the cerebral cortex and there's the cerebral cortex from a map by Burkhalter's group again. And here are all the areas in the cerebral cortex of the mouse flattened out into two dimensions. But the nice thing, there are many nice things about the mouse brain one of the very nice things for imagers such as myself is that it doesn't have cell cyan gyri, it's listencephalic or flat. So one can look directly at a broad expanse of the visual system and I like using again the metaphor of astronomy if we build observatories where do we point them we can do both a wide field survey and there are all sorts of synoptic surveys of the night sky the Sloan Digital Sky Survey and others but there are also the narrow field observatories the Hubble Deep Field and I think Ultra Deep Field very long exposures from the Hubble Space Telescope looking back near the beginning of the universe more than 10 billion light rears away and showing again a narrow field of early galaxies. So what's the metaphor there? The physiological imaging of multiple cortical areas is wide field imaging where we have a quite large window roughly a fifth of the mouse cerebral cortex which encompasses multiple areas most of which are visually driven and what we do is we do this type of calcium imaging which is a good measure of the spiking activity of neurons and we look this is before virus injection the virus holds a genetically encoded calcium indicator after a targeted virus injection we can look at two in this case of the multiple visual cortical areas and then in order to look at single neuron resolution we zoom in to fields of view that are 100 microns to half a millimeter or maybe a millimeter in scale and when we zoom in we see a field of view that looked like the movie I showed you earlier where this field of view has in this case 100 neurons that we can study the firing of those neurons with calcium indicators and find out for instance which of these cells respond to vertical stimuli which respond to horizontal stimuli which of them are modulated behavior which of them are not in this case the color code encodes the responses to the different visual stimuli of different orientations so the green neurons are color green because they respond to vertical orientations the red neurons are colored red because they respond to horizontal orientations and the blue and the yellow are the two obliques and that gives us a microscopic view of what individual neurons do but we can zoom in yet further to do a very fine scale anatomy at the EM at the ultra structural level and look at again serial section electron microscopy section after section with section thickness on the order of 40 nanometers looking at XY resolution of 4 nanometers and we're at that resolution we can see every axon dendrite and synapse and then we have the extraordinarily hard job that you heard about this morning of segmenting the data but that's again a problem for the audience that neural computation, neuroinformatics one of the grand challenges is analyzing these large data sets and that's what I'll be telling you about in the next number of slides and just stepping back way back to the issue of the structure of cortical networks of course there are two types excitatory inhibitory cells for me probably the most important fact by the excitatory network is that it's sparsely interconnected with neighboring neurons of the same cell type right next to each other the probability of connection is at most 10 maybe 20% connected and that poses the important question which neurons are connected to which and how does that relate to the computation being performed by those neurons so there are dominant models that the excitatory network is sparse and random and there are models that are actually yet again becoming dominant these things go back and forth are there cell assemblies are there within any cortical networks are there sub-assemblies of cells that are preferentially interconnected and those might be cells that do the same thing or more simply and more generally neurons that fire synchronously and any heavy network neurons that fire synchronously tend to be more connected but it's a remarkable and tumbling fact of current neuroscience is that although we believe in these plasticity rules we really don't know what is the end result of these plasticity rules how is the cortical network wired up how does that relate to co-activity and how it relates to function those are all just they remain just as much of a black box as they were in the 60s with a few notable exceptions but close so the general question of the functional conectomics part of the neural coding program is what do individual neurons do that's functional imaging and how are they wired up with respect to what they do and for instance to the vertically oriented cells connected the vertically oriented cells to the cells that signal fast velocities versus slow velocities are those interconnected and neurons that do vastly different very different things not interconnected and that's again not one question about neuronal function and it's not one question about connectivity it's are the feed forward connections from the thalamus rule-based from layer 4 to layer 2, 3 are those rule-based the recurrent connections in layer 2, 3 and the feedback connections are all of these what are they rule-based and if so what are the rules and in order to do this we have been doing large-scale electron microscopy correlated to physiology so remember I showed at the very beginning a zoom in to a red and a green cell and asked the question how are they preferentially interconnected this is the actual EM data from those red and green cells we knew what they did in vivo but in fact it's data from a very large array of cells it was a project led by Davy Bach and Wei Chang Li in my lab at Harvard and it's being continued by Nuno de Costa and others at the Allen Institute and also by Wei Li at Harvard correlating function with fine-scale connectivity with electron microscopy but in order to do that you need a lot of electron microscopy here's the cortical surface of this EM image the bright things or cell bodies that you can see pretty well the very brightest things are blood vessels and in the volume that we looked at you can almost think of it as a cortical slice it has 1500 cells on the order of 10 million synapses and its spatial extent was 400 microns 130 microns by 350 microns each picture was pushing 10 giga pixels there were 1100 of these pictures and when I say picture these are all montages the actual number of pictures from the EM system that we built it was 3 or 4 million individual pictures that were all stitched together in three dimensions by a group at the Pittsburgh supercomputing center that were put together or more terabytes of data we needed high-throughput imaging and a lot of post-processing so if there's any neuroinformatic problem this is one it's tera-scale and in the new the new project that the data of the way we connected at Harvard the raw data were well more than 100 terabytes so it's 0.1 peta-scale individual imaging data set then the question is what do you do with it and first of all you zoom into it so you can see the individual components here's a 3D view of the two cells that I showed at the beginning but that doesn't show you very much other than the nuclei and the apical dendrites but if you zoom in and zoom in again I love big screens this is maybe 3 meters so at this scale the entire data set is three football fields by three football fields of data just in 2D times a thousand so it's just an infinite data set but the point of it being so big is that you have to have a very large extent to encompass cortical neurons but you also need the nanometer scale resolution to see what you really care about which is first the cell body and then the hardest thing are the little axons these little axons, the cutting cross-section are tubes cutting cross-section that you trace through multiple images and there they can be as small as 70 nanometers so those are seven little axons that need to be traced by computer programs or currently through beleaguered undergraduates they seem to enjoy it by human tracers and the goal is to trace axons until you see a synapse jump that synapse go back to the cell body and then you have a connection from a cell body through an axon, through a spine through a dendrite to another cell body segmented all by hand in the first study that we published a couple years ago very importantly these data are online in the open connecthomeproject.org Randall Burns told you about this two days ago and it's a really wonderful public resource for this type of data of course the Allen Institute when we get into this business we'll host the data ourselves as well but currently it's not computer programs doing the tracing but an enormous amount of work that here we have some number of cells that we characterize physiologically and then with an enormous amount of tracing these are the postsynaptic targets of those cells I won't tell you anything about the science of this project but I'll tell you about the goal of the project in general the second generation and future generations is again is to ask pose this question how does connectivity relate to function? five minutes and the most important question that people always ask is why do these unitary experiments want to be different every time but the goal is not to find the details of one brain circuit but to find the rules by rule I would say how does connectivity relate to function and motifs can be independent of function are there highly interconnected sub networks that would be a motif and like any part of science looking for the regularities rather than the specificities the second so that's functional conectomics the next part really the final part is functional projectome where a projectome is not a point to point neuron to neuron connectivity but an area to area connectivity and again there are two aspects of our endeavor one is to look at cortical computation within a single cortical network and we're using primarily primary visual cortex and the second is to look at the network of networks multiple cortical areas and this science comes from a couple of studies from my lab at Harvard but studies from other groups as well here we have primary visual cortex and five of the nine or now eleven different higher cortical areas and the first question one can pose of these areas is what do they do differently V1 if we understand the receptive fields and how those neurons relate to behavior what do the higher visual cortical areas do or more simply what are the receptive field properties characteristics of these neurons in different visual areas and a couple years ago our group and Ed Calloway's group published surveys of response properties in different cortical areas and there are a lot of details in those studies but an oversimplification is that primary visual cortex responds to a broad range of different stimuli in particular responds to a broad range of stimulus velocities over a factor of more than a thousand different velocities and the higher visual areas some of them respond to a broad range and some of them are very specialized to high or low velocities and also finer or course structures so that's what these early studies did a more recent study that came out this year from Lindsay Glickfeld is how does primary visual cortex distribute the information to higher visual areas what are the rules of projections between different cortical areas and the hypotheses again it's a whole string of hypotheses but these hypotheses have the form do functionally distinct populations of V1 neurons project to different cortical areas and they can be functionally distinct populations that respond to different velocities or they can be different cell types and that's where the whole mindset project where we can look both at function and cell types in different areas there are a large number of questions about cortical networks that we can examine with this type of technique and the question is do they match the functions of higher cortical areas the technique is very similar to the connection atlas or projection atlas that the Allen Institute is already performing make an injection in V1 a virus that makes neurons express a fluorescent protein in this case it's a fluorescent calcium indicator so here we have you can see in the different cortical areas weaker fluorescence than the cell bodies give but significant fluorescence of the projection from V1 to different cortical areas and if you zoom in I showed you this movie earlier this is one of these things Lindsay first started doing these experiments and you can just look at this the kind of experiment that at least made us smile the early experiments a decade ago when we could see visual cortical single neurons flash when they responded to different stimuli now we can see single axons and even individual butons this is sped up but look right here right above where the pointer is pointing there are two butons and sometimes they fire together sometimes they fire at different times there are clearly two butons from different neurons firing away and we can look so these are neurons in primary visual cortex in a higher area and we can study the receptive field properties of these neurons so we can I could go through the details but obviously there's no time we can look at the speed tuning of these butons and I told you in V1 different neurons have a broad range of speed tuning some respond only to very slow stimuli and some only to very fast there are two cortical areas AL and PM where if you look at the cell bodies the cell bodies respond to either fast stimuli or slow stimuli and it would stand to reason but it's actually it doesn't have to be true that the projections from V1 to these higher areas match those areas that the V1 neurons that project to AL have speed tuning so that some of them respond to 400 degrees per second or even 800 degrees per second but and very few to 1 degree per second however in PM neurons that respond to 1 degree per second or 6 degree per second are much more V1 neurons are much more likely to project than neurons that respond very quickly and there's a V2 like area and the projection from V1 is very general so again for this one projection we know that neurons V1 have diverse properties and they project to distinct areas but that's an anecdote and neuroscience is full of anecdotes our first physiological pipeline in the neural coding program is the functional projectome and this will be similar to the connection atlas but with physiology and it's a very large matrix of experiments that each one of these squares is a distinct injection and the injection will be in one of 11 areas or thalamic targets projection cell types it'll be in one of several cell types a layer 5 cell type a layer 2-3 cell types each of which will have distinct projection patterns so that's the anatomy experiment and then the physiology experiment is again multi-dimensional we'll look in N different target areas you know from V1 to 10-11 different target areas we'll have different stimulus types velocity tuning noise stimuli to map receptive fields natural scenes etc and of animals in different behaviors running doing a cognitive task and sleeping there's a lot of neuroscience in there that we hope will be used by people in a reverse array of fields and again it's neuroinformatics and really our goal is to get people interested in physiology and anatomy and their interrelations munching and doing science under a very large data set now the goal I'm almost done I told you about the mouse cell types program very briefly I told you in great detail about the neural coding program and two large scale projects in that the functional connectome, the functional projectome and then very briefly apologies Christoph and Mike I'll tell you about the modeling and theory component but again it's an equal component and really it brings the synthesis together there are component models, system models and theory the component models are generalized linear and nonlinear single neuron models adapting synaptic models sort of different types of single neuron models system models spans a very broad and very important range ranging from biophysically realistic models of the type that in particular the human brain project is working on quite a bit more reduced models, generalized integrated fire models of v1 and what Christoph and the team call IC a generalized integrated fire point model of the entire visual system from retina to thalamus to multiple cortical areas and finally it wouldn't be leading to understanding unless there were theories, so theories of dynamics of Bayesian coding and reconstruction and hierarchical models such as Hmax of cortical computation and the near term goals of the modeling this is the next slide the near term goals of modeling and theory again are concentrating on primary visual cortex and in fact even layer 4 a model of the optics of the mouse eye the morphology and just the numbers of cell types in v1 and their interconnections spiking models for the thalamus and the cortical targets spiking models of an entire cortical column and how it relates to both receptive fields and population statistics and dynamics and finally large scale models of the network of networks from the connectivity atlas I guess and also this is not modeling but analysis and there is a wonderful computational anatomy team led by Hanh Chuan Peng and I told you about that so in the last slide really I'll talk about the general project of Minescope and its challenges what we have are plans for a very large scale project by the end of the decade there will be 250 people working on the Minescope project in an institute that will be twice that size an easy to state goal is to build these observatories of doing large scale experiments and that's something that has experimental challenges, informatics challenges but I think the more important part of this large project is the tight integration of anatomy, physiology, modeling and theory and the real goal is to have groups to achieve synergy so our watch word, our journal it's called synergy where the entire group ranging from mathematical physicists to intracellular physiologists to molecular biologists get together and discuss the problem of coding and computation in the visual cortex and really the goal, the related goal of making all of these scientists who have distinct interests but have signed on to work on one common project, I think that's really the hallmark of this project at the Allen Institute it's one institute, one project one goal with a large number of people targeted on the same thing and it's really as Kristoff and we all say it's as much a scientific project as an experiment with the sociology of neuroscience and really this is the important part of the team that this is the team a year ago before I think any of the mind scope people have arrived and you can see how large it is there, it's over 100, now it's well over 200 and we'll be growing to 500 by the end of the decade and it's really just an extraordinary time in neuroscience when these large scale projects are happening at single institutes and nationwide or continent-wide here in Europe and it's remarkable to me and really remarkable to the institute that the generosity of two people, Paul and Jody Allen allow us to do this comparable scale project in neuroscience thanks Yeah maybe I can start yeah, so first I think it's a wonderful talk and a wonderful project and it's sort of like the interesting thing is that the institute which is most open to data sharing in practice is actually privately owned it should be the other way in some sense or privately driven, privately funded it's scary for individual groups to do it but if it's the price of doing business it's the privilege of doing business is to have the data curated and put online so I think many people here in this room would like to use some of this data that comes out, so what's the time scale for getting out morphologies of the neurons or maybe like multi-compartmental models of the neurons how do you see, when can you expect to have them online? No, frankly the original plan was to not have any of the mind scope pipelines ramping up until 2015 we've been somewhat over eager and we'll be starting the functional projectome and a morphology and biophysics pipeline and also the next generation connectivity pipeline next year, I we're planning and this isn't a promise even though it's public I guess on YouTube that by the beginning of 2015 the first data sets for mind scope will be going online Hi, I'm Michael Bota from University of Southern California can we go back for just a moment to the Ponectom slide where you show that great matrix I have a question, let's say with two parts and it's better to have the matrix You wanted the matrix of the mouse connectivity atlas? Okay Okay, so these data where from crane injections or classical tracers? They're AAV injections AAV so overwhelmingly enter a great connections and what do you think what are you going to do with retrograde or labeling? That is one component of what we called next-gen connectivity so stay tuned Great And again, this is Hank Weissang's project The second part, let's say the second part of the question you mentioned that you have quantitative data in this matrix did you actually count the terminals? No It's overall fluorescence waiting for more informatics but that's what it currently is Intensity, optical intensities, something like this Optical intensities Optical intensities So how are you dissociating between a fiber of passage and a terminal? When you say when you have overlap of these two in a workshop? No, it's a concern This is the detailed analysis but it's not that's why the data are online The beauty of the Allen Institute is that it works with a timeline Is it the final word on this project? No, but is it the definitive first word? Absolutely You had to produce something by a time It's more than that So is it the most quantitative analysis of the point-to-point connectivity? No but I think and they think it's not my work signal intensity is something like this It's closely related Would you be interested in some other people to help you annotating and polluting manually this data? Absolutely These data are extremely valuable They are unique The data are all online There are means for getting the entire data set It's a header byte so you'd have to pick and choose but it is online and the institute will be continuing to work on it but of course the community would be a great feature There would be ways that the community didn't do further analysis Thank you very much