 1990 under the term mapping the brain and its functions integrating enabling technologies in a neuroscience research with Joe Martin as the chair and co-editor of the report Constance Pachura and You can see here that there was a long list of representatives of the public and the academic and private Domains and Sure enough here's vent surf Representing the new age of the internet. I Had the privilege of co-chairing a subcommittee of the of the committee with him on the diversity of neurons which were impressive and and to him and I Made it clear to him and everyone on the committee that Neuroscience data are much more complex than sequence data The genome project informatics is trivial by comparison One of the important things that the committee dedicated itself to right from the start was that all Information needed to be part of this What was conceived of is this? Integrated map of the brain and this represents what Sean has already shown you the levels from genes and proteins up through cells and circuits to behavior And I think that was a key for why we are all gathered together here to pursue all that range of disciplines and data that Sean told us about Illustrated The some of the key Recommendations of the committee where that there should be a long-term goal of Multidisciplinary multi-level mapping of the brain that there should be the development of integrated and Interrelated databases. I'm just Extracting here from the recommendations specific recommendations and finally that in rather than having one big giant monolithic program that a Static strategy should be followed of pilot projects dealing with the diverse data by actual R01 type program projects projects individual projects The I think it's appropriate to recognize The initial leaders of that human brain project That started us off Steve Koslow the founding director along with the co-director Michael Herta We opted for the title human brain project To emphasize the vision that we had that would be comparable to the vision of the human genome project But it became clear That I the real Challenge we had was to develop the tools in order to reach that goal of an integrated human brain project and So the term in neuro informatics Began to emerge and I thought it was interesting to try to find out when the term actually Evolved the earliest I've been able to find Was the establishment of the Institute of neuro informatics quite separate from the human brain project of course at the at the University and the ETH of Zurich With Rodney Douglas and Kevin Martin in 1995 There was then a volume By Koslow and Hurt on entitled neuro informatics in 1997 We did a review article for trends In 1998 that used the term And then there were following volumes that used the term as the as the overall description of These various tools and databases that were beginning to emerge We can discuss later Others who may have Sight know about citations that are earlier than this So an important part of the recommendations of the original committee was that diverse data should be gathered and integrated in terms of computational Quantitative computational models And that then meant that The rise of neuro informatics would be closely associated with computational neuroscience computational neuroscience had its origins of course in the Hodgkin and Huxley action potential model in 1952 That was the 60th anniversary of that publication those Five publications was held in Cambridge, England just earlier in the year At that time We also Were part of that celebration because Wilfred Rohl then introduced Compartmental modeling in the 60s as represented here the compartmentalization of branching dendritic trees plus the representation of synaptic potentials and so from those two Publications we date the origins of computational neuroscience. I joined Rohl in the 60s and together we applied a Hodgkin-Huxley like model of the action potential together with the compartmental analysis approach of will To olfactory bulb neurons and mark has mentioned that this led then to a compartmental model based on electrophysiological recordings that predicted a dendrodendritic Synaptic interaction between mitral and granule cells in the olfactory bulb that predicted these actual Side-by-side Excitatory and inhibitory synapses mediating lateral inhibition of mitral cells to the granule cells So as Mark said this could well be the first Example of a micro circuit It includes of course everything from the synaptic level to the dendritic level and To the micro circuit level so it represents the sort of goal that that was envisaged envisaged in the original human brain project meetings of integrating data into computational models and In the succeeding years we built that model into the sequence of Stages of Processing that take place in the olfactory system from the initial transduction of the order information contained in order molecules in the olfactory epithelium by the olfactory receptors to the representation of that information in activity maps in the glomeruli the processing of those maps by the mitral granule cell lateral interactions the construction of a content addressable memory in olfactory cortex and on to orbital frontal cortex Within the prefrontal cortex the highest level of the brain Where perception is carried out? so to support that work we developed a number of digital tools and In our review in 98 we Summarized some of the tools that were needed in general and that we then proposed to develop in specifically and that led to the construction of what we call sense lab a long-term effort to build integrated multidisciplinary models of neurons and neural systems that includes several databases related to the Specifically olfactory problem, although it has a general interest because the olfactory problem starts with the largest gene family in the genome the olfactory receptors of which there over a thousand in virtually all mammals and so we In the early efforts to sequence the Genes by several laboratories in the 1990s They came to us and asked if we would construct a database that would support this community effort and starting with that we have added to the Sequences that have been generated over the years and there are now over 14,000 Receptor sequences in in the ORDB and if there's time I at the end I could come back and I show you How that database works? But I also have a demo where I can also show you that for those who are interested We also have a database of the odor molecules interact with the receptors and An odor map database of the maps that are generated representing the information in the order molecules But I thought there would be more interest general interest in this group in our neuronal databases that start out with a database that represents an inventory of the receptors and channels and neurotransmitters that are found in given cell types With then a connection to neuron DB which represents The expression of those properties in different parts of the neuron then how those are incorporated into models of both for Channels and neurons and finally into microcircuits So it's here that I go on the high wire and Let's go to a browser and I thought that the essence of databases and digital tools is how they're actually used at the desk or In the lab when you've got a few minutes or a few seconds to look up something And so it seems to me the key to the tools that we're developing is How accessible and how quickly usable they are so I'm going to try to show you that now I May only demonstrate that it's not feasible in a lecture hall, but let's give it a try so our first database is cell prop DB and This is a resource That is intended to serve as a repository for data on gene products expressed in different brain regions Support research on cellular properties gateway for inputting data into canonical neuron forms and neuron DB Identify receptors across neuron types to aid in drug development Serve as a first step toward functional genomics and the teaching aid so we can go into the The database and we have a number of cell types represented our interest Of course is primarily in the olfactory bulb But almost as much in comparing the properties in the system We're working on with properties in other neurons in order to carry out a comparative analysis that gives a context for understanding any given system and I think that comparative context is one of the main themes. I'd like to get across today and also With the philosophy of sense lab. So here we have the olfactory bulb mitral cell and There's a representation of it morphologically and here's an inventory of properties That highlighted are the profile for the mitral cell And so you can see the receptors currents and transmitters Are represented here and so let's say that we are interested and so this gives the profile for the mitral cell now, let's say that we're interested in Comparing the mitral cell with GABA Receptors with all other cells that have GABA receptors So we can go down here and click on that and this immediately brings up then a subset of cells that all have GABA receptors on Them and so you can see that this includes both principal neurons output neurons and various kinds of inner neurons as well across many parts of the nervous system Some of which one may be aware of like the Hill campus others not such as the retinal ganglion cells or cells in the dorsal cochlear nucleus We can in addition to a single Property put in a variety of any any combination of properties that we think might Represent some significant kinds of activity and so under currents we can we can ask for transient sodium currents and a currents for example and Go to submit those and this immediately then gives us a somewhat smaller subset of cells now that are Expressing that combination of properties and so this I You can see that this could Be used for any particular set of properties that you're interested in but I think the main point here is that we're beginning to move toward a multi disciplinary multi-field Approach to Identifying families across all of a particular a set of Entities in a database so like in a gene in a sequence database you can look for particular combinations of sequences and Identify new classes and families of receptors or channels in the same way we can do that for different types of cells in in the in the nervous system however It's obvious to most of us who work on cells that it's not just The expression of different properties by a whole cell But it's where in the cell those properties are expressed because a cell contains a number of compartments virtual compartments In which integrative different types of integrative action occur and so for example for the mitral cells we have proximal region a middle region and a distal region of the Primary dendrite. This is the part of distal part out in the glomerulus. The same can be said of the lateral dendrites the soma action axon hillock axon and Terminals complete the sort of basic compartmental structure of a neuron. We call this a canonical representation of the neuron it enables us to identify the major integrative sites in a cell But it also means that if we do this for all cells we can search across Different compartments of different cells and be much more specific about this new classification way of classifying cells So we then go to our next database. We're right there in neuron DB And so here's that same cell and now here's the canonical representation of it in a compartmental way And now we can bring up the data and now instead of being for the entire cell This is just for the distal apical dendrite and now we can ask And so first of all we see that this is the integrative structure for this compartment and it's more than just a Gabba receptor or glutamate receptor. It's also the intrinsic currents and in this case It's also the fact that this compartment can release glutamate It's presynaptic as well, but let's now ask our same question How about all distal apical dendritic distal compartments of dendrites that express Gabba Gabba receptors and here we have now a subset. So you see it's a smaller subset than we saw with the whole cell Representation now at a compartment at a distal Dendritic compartment level. It's a smaller subset and again it crosses both principal neurons and interneurons So now we can ask the same question that we did before of adding in other kinds of intrinsic currents So let's ask now about our transient sodium current and our a current and now you see these have a completely different significance Because the transient sodium channel is of course present in many most neurons that are generating actual potentials in their axon But now we've got a distal apical a distal apical dendritic compartment That we're also asking does it have the ability to generate action potentials and are there also a currents? Which usually are present in order to balance out that excitability and so now if we search We see that there that there are that there are in addition to the mitral cell There are the ca1 pyramidal neuron the dentrate granule cell cerebellar-pikinji cell and the neocortical pyramidal neuron also have that this particular combination of properties and now it gets it's getting interesting because that means that Active properties can be involved in the integrative action the activity that's taking place in that distal compartment Now at this point you might be you might be interested in The fact that the ca1 pyramidal neuron is also part of this subset So let's just go and see what it looks like and immediately. We're in the ca1 hippocampal pyramidal neuron and you can see it has its own set of Properties and we could fill in all of these and look or many of them and look for different combinations across this This are other types of neurons. So that indicates the kind of Exploration one can make of the integrative structure of every nerve cell in the database and In a within a comparative Context of relating it to any of the other neurons in the database and bringing out new classes of neurons that have particular combinations of properties in addition We build connectivity into this so that we specify the connections that are responsible for the inputs to the receptors and For the connections made from the output from the axons or from the dendrites and in the in this in this particular case And so we are building in then what can become the basis for the connect to them a micro connectome of these cells and Also within a comparative context of other what one can call micro connectomes and then finally we provide annotations of classical references for the first demonstrations of these properties in order to establish the They're defining characteristics Okay, at this point we might say well We're interested now. We see that there is evidence for these different kinds of receptors and currents We see some of the classical references to it What kind of models have been built based on these? experimental properties and we can go immediately then to model db and Here for the mitral cell Is a list of the different kinds of compartmental models of a mitral cell and I just show for as an example a Model that we did a number of years ago with Michael Heinz in the of the mitral cell and This indicates how one can go to the actual files for running that model This is for the transient sodium and The key the first key is that you can run the model immediately. You don't have to build this model From the publication as one used to have to do one can actually take the model as it was Run for a paper and immediately run it yourself So it comes I think as close as we can get to the way that every publication needs to provide the specific information about the agents for Obtaining those results so that they can be repeated and that's a key for computational Neuroscience is that we've got to be able to repeat those experiments by other labs In order to build on the finding and the key here in the next key is that each of these properties can be swiped and changed so that this model can become your model if you as you Begin to run it and have other info have other kinds of data that would provide you with a way of running it in a new way so we can also then go to model DB the home page and explore the Inventory of the models in many different ways so by model name We can see that under model name. We have 734 models now each of them curated by Tom Morse in the lab you can also explore them by first author each author or by Region here is by cell type and so you can go in and find a Model in your particular region that you would be interested in Here are by transmitters and these are the different kinds of cells that are GABA ergic or or Miturgy or whatever we also have Different simulators so although We have our own interest in the simulator neuron because Michael Heinz is one of the group and has maintained neuron for well over 20 25 years We have a number of other simulators that are represented you can see here There are over 70 different simulators Matt lab is well represented Genesis is well represented and so and so forth so That that illustrate indicates that whatever a simulator you you use Will will probably be represented in one or another of these these models We can also search by Author so one of the classic models is oopies of the olfactory Mitral cell as well as others here So therefore by him and you can go in and examine those as well So this illustrates then the The aim of getting of using models to integrate the information in the as represented in cell prop and in neuron DB into these quantitative models for For each each cell type and then finally we go to micro circuit database and here we have Started to build up the models specifically representing micro circuits and here we have Something like 180 or so of these kinds of these kinds of models that are represent both realistic micro circuits the kind that are built on realistic representations of the neurons as I've shown you in neuron DB and also connection as networks that are of the artificial neural network type Okay, so I think we're in reasonably good time here So let me now Let me now go Hold on. I think we want to get out of this and go back to there So now we are back in our PowerPoint So now I'd like to show you several ways that several of the kinds of results that have been produced by this Using the information that's in sense lab one is a paper that McKelley, Miliori, and I did number of years ago in which he Made a preliminary effort to relate the expression of different. I am ionic channels I Relate them to the different morphologies of different kinds different kinds of dendritic morphologies and so starting out with synaptic inputs activating Cells with large or small temporal windows he was able to work his way through a Decision tree that enabled him to identify the sequence of decision points that produced in the expression of properties that produced the properties of the different kinds of cells Including mitral cells neocortical and ca1 cells and Phthalomocortical and Pekingia et cetera on this side. So again, that's a new way of classifying cells in relation to integrative properties That is that cuts across the different kinds of morphological types that we tend to be very very much fixated on And at the microcircuit level we with McKelley and with Tom McTavish and Michael Hines have been Developing a scaled-up model realistic model of the mitral cell granule cell interactions And this represents current stage we're at in which we have 500 mitral cells receiving input From glomerular activity patterns is determined by Kinsaku Mori's experiments using intrinsic imaging Interacting with 10,000 granule cells and this is a site as shown by Mori's Maps where there is a relatively weak input to a set of mitral cells and this shows over time the emergence of that excitatory activity with the emergence of inhibition Ladle inhibition on either side due to the activation of granule cells starting at this at this point in the learning process this is a much stronger input giving you much stronger lateral inhibition and So forth and the key here is to show how extensive Ladle inhibition is within this region of the of the olfactory bulb much much more extensive than I've been previously thought but as is going to be needed in order to explain these distance independent nature of the processing that takes place within the olfactory bulb There's a lot of interest now in systems biology and in systems in terms of Different systems in the in the nerve system What we have in the olfactory bulb is a need to recognize that those systems are Controlled by the respiration of the animal in delivering the sensory stimulus to the sensory receptors and here we're this is a Recent experiments published by Matt Phillips In which the lateral inhibition of the mitral cells differs Depending on the phase of respiration whether it's inspiration or expiration and you can see that lateral inhibition is sensitive to the phase of the of the respiration and so that means that our systems biology Has to take account not only of the neural systems, but also of the body systems that are providing the input and so That I remind you that all of this is aimed then at this sequence of operations that's taking place within the microcircuits of this of this system from the olfactory bulb olfactory cortex orbital frontal cortex and so here I think I'm going to go back to the to the high wire because I'd like to now and By indicating the context for sense lab in relation to Some of the other major efforts that are taking place in order to develop neuroinformatics tools and to move toward this integration and the main theme here is that INCF and neuroinformatics has Now focuses very largely on tools or predominantly on tools because it's been such a struggle From the start with the human brain project to develop the tools in order to work toward this goal of Brain integration and in fact it was a problem with the human brain project During the early years of funding because people said well, where's where's the result? and in fact it was a much of the effort was put in by the Brain imagers for example just warping the brain and being able to compare the results in different labs and I remember when Mike gazaniga as editor of the Journal of Cognitive Neuroscience I Suggested a policy in which all papers submitted to the Journal of Cognitive Neuroscience should provide the data sets for the generation of those Results there was a tremendous outlaw from the Field because they were reluctant to do that and it was only after a couple of years when a review committee gave Strong support for this idea that there was a general recognition that this is the way you share data in that particular field You share your data sets so that people can see how the data was generated and can do their own comparisons And so I think there have been so many issues about building databases and Building the tools that we've lost Site a little bit lost site of what the real goal is and that is the integration and So I'd like to give an indication of how we might move toward that in that integration In terms of the tools will need to do it and the integrative approaches so Let me Start out here So one of the things we need to represent we need to have a general Recognition of What the resources are that go beyond the particular ones that we're building and One of the ones that I think is important to know about is the IU FAR no manclisher. How many people know about IU FAR? Okay, few people so here's here's for the rest of us I I would probably be among the ones who wouldn't be so aware except that I was put on the committee to to represent all factory receptors it turned out that the number of all factory receptors is greater than the number of all the other gprs CRs put together and They have developed a nomenclature for all the others, but it We found that it was just not possible to apply it to the olfactory receptors so it's the illustration of Some of the problems in developing general Nomenclature that apply to all parts of the system. So here for example our G protein couple receptors and Here for example are the five hydro hydroxy trypone receptors and if you go into any one of these You see you bring up a Tremendous amount of data here Tremendous amount of data about agonist antagonists Transduction mechanisms tissue distribution so forth. So this is definitely a resource that we ought to know about the same The same is available with regard to ion channels. So here are sodium channels for example here are One kind of sodium channel and again you go down here and you can see the very rich amount of Information there is annotation and all the rest about any given channel type so again, I think if we if this level is going to be a A Essential part of building up the the multi-scale Integrated view of the brain that we're all after then we need to have be aware of the Resources that are going to be adequate to that so the next I'd like to show is the NIF So Now if you key the NIF in You'll get the national ignition facility and there's no way we can get rid of it We're working hard. We're her working hard to displace it but And now that we're in now that we're in Germany. We can't even get up to As high as we really ought to be but nope that's that was the wrong one So how about this one here? Here we go So here we are with the NIF and I assume Most most of you have heard about the neuroscience information framework led by Marianne March on She's here yet but the This is a portal with a tremendous amount of Functionality built into it and you'll hear more about it in later in later Talks, but I thought I what I would show you is something that we're involved in and that is if you go down here to Neurolex and And go to neurons in the hierarchies then you get a list of neurons of Subcategories of neurons is the neuron types That is now up to over 200 around 250 So this is a much more comprehensive. This is a truly comprehensive attempt to Characterize the main types of neurons throughout the nervous system. It will include both vertebrate and invertebrate and so I the One of the examples I wanted to show of how this works is in the amygdala here So here is How each of these entries we're aiming to provide for basic information and then a minimum of Specific properties soma dendrite axon specific properties and intrinsic properties And so you see this is similar to cell prop But it attempts to identify properties that become the defining properties of a given cell type And at the same time to enable us to go in and for example with GABA to identify All of the Cells that use GABA This is one way of representing it, but there's an even better way and that's to Go back and so you you could ask what are what are all of the cells that share a given Neurotransmitter As the transmitter released and here quite quite quickly you can go and get an entire Inventory of all those 250 cells For which ones are cholinergic? Which ones are GABA ergic? Which ones are glutamatergic? Etc. And so we want to do that now For all the other properties that are listed there. So give me all the spiny neurons give me all the paramedal neurons Etc. And Working with Steven Larson. We have a small grant from the from the from INCF to engage domain experts in helping us to Identify these main cell types and their properties in different areas. I just show you the curator page here And you can see that we already have some 20 people Who've signed on to provide this information and we encourage anyone who would like to volunteer For their own favorite neuron to to let us know So the next I'd like to show is The Ellen Brain Atlas and I don't think I need to Belabor this because they'll be I'm sure a much discussion about the Resources available in the Ellen Brain Atlas But one of the things we want to do with neuron with neural X neuron and with cell prop is to link to the the different patterns laminar and otherwise of staining and localization of properties So that one can go from those properties to the individual cell types in a particular region or layer or vice versa Another another project that ought to needs to be part of this Overview is the blue brain project that Sean Hill is involved with Henry Markham in running in Los An and I think this is also well known as an attempt to provide a comprehensive description of the court the Cerebral cortex in terms of these different layers levels of organization that Sean was illustrating the blue brain project is now a Core site for the proposal for a Blue brain for a human brain project. So now after the blue after the human brain project ended in 2000 around 2005 in the US INCF came in and now the the blue brain project Came in and now the human brain project and so that has its own website as well And I thought I should just indicate I think some of you may have heard that this is a large very large program that they're competing for This is not a program in biology or in neuroscience the human brain project you Sean can explain further is Applying with it for funds within a very high technology cutting-edge technology Grant program. It's the I think the only one represented in neuroscience because they proposed to use cutting-edge computational and other other Methodologies, but the the research areas here are very impressive in bringing together these different Areas of research that I can't I can't get down Here we go So you can see that there a whole range of issues is going to is going to be included in this trans European Project This again represents This original goal of the human brain project of doing the actual integration and it shows how Comprehensive this kind of integration is going to need to be not only for the the scientific aspects of Brain organization, but all these other aspects of it the medic medical the supercomputing aspects the educational aspects and the ethical legal and social sciences aspects Finally We don't want to forget the connectome project and How many here are part of the connectome project? Not so many But this is a tremendous project in the you mainly focused in the US For carrying out a comprehensive analysis of the human Connectome and so I I don't know whether we need this big a program tens of billions of dollars to do the the to do the Macroconnectome, but we certainly I think want to think in terms of a micro connectome project That would connect with the macro connectome project because after all why do we have a macro connectome? It's to connect the parts of the brain. They're actually doing the computation So let me end by Going back to So let me let me go ahead here. I think we were here Okay The future So what I what I would propose is that we need to work toward our Principles of comparative neural systems, so we each work on our own neural system, but I think comparing systems Across other parts of the brain comparing systems across other Species I that have and the system we're working on and comparing across Birdabrit and invertebrate as well. Here's an example of what I would call the comparative approach I did this study several years ago In the olfactory sister in the olfactory cortex. We've identified a basic circuit in which the input comes into the apical dendrites and the apical dendrites then the Parameter learns the parameter learns and have recurrent excitatory collaterals recurrent inhibitory Circuits feeding back on them with some feed-forward as well We propose that this is a basic circuit and indeed you can apply this in with minor modifications to the hippocampus and to Turtle dorsal cortex as as Kriegstein and Connors did One can then propose that Neocortex arose out of a kind of a duplication of that organization so that one has superficial and deep Divisions of the Neocortex layers 2 3 and 5 5 6 particularly in which and their this basic circuit is embedded in each one This has been used for example Recently by Fritschoff-Helmschen At a genelia meeting and we attended in which he showed an adaptation of that to show How it might help to account for the results that he had got in identifying different In kinds of inputs coming in here from the VPM here from the posterior medial Here from other areas of cortex So that this provided a way of thinking in an integrative context about the about his own analysis of these pathways in the cortex here is an example of the Kevin Martin Rodney Douglas Neocortical Canonical microcircuit of which has two three and five six with recurrent Excitation with current inhibition in very similar ways essentially the same circuit that I showed you and so I think you can see that there are indications of a convergence of the integrative approach And I think we should continually look for these convergences as ways of beginning to build a Consensus representation of the map and so a brain map and so I leave you with the question a human micro Connectome project is that in the future for us? in which the micro macro connectome connects the the micro connectome and that maybe this is a A Goal that the NIF can Provide a driving force for as well as the INCF And finally this is the sense lab team Perry Miller has been the backbone of our informatics Michael Heinz Has been the backbone of our computation and all these other Colleagues have been just a wonderful group to work with. Thank you very much coffee. Yes. Yeah, sure questions In the part of your talk where you were discussing the computational models in your your databases and keeping track of The models in a way that they could just be in sort of installed and run right and I wonder if you could say a few words on Challenges in doing that in sort of making sure that you're keeping up with the literature on that So this is a very good question. So if you make models easily available for running, I think what you're saying is if anybody can run the models maybe they don't have the background in understanding the the nuances of the data this is one of the problems of elect physiology Particularly that you really need to have a thorough understanding of the properties in order to begin to build your own representation of a Particular model is that is that what you're doing? No That's a great answer, but it's not my question. So my question is so you've got a real if you're doing that Which I think is extremely valuable You've got to kind of keep up some find a way to keep up with all the new models that are being published in the literature And isn't that challenging? Well, that's that's sort of what I was getting at Oh, okay, if you don't if you're not familiar with the field then are you running your model in In a field in which you have no you have no particular background Is that what you're saying and so you as a curator you as curating the database How do you keep the database current with the new all the new models are published or maybe I'm Overestimating the volume the database the database is accumulative So all it it starts it goes back to Classical models like made in the Sonowski in 1995 and it's there they're all there and you can Run them or you can update them as as you wish and You can also search you could also Determine when you search them which are the latest ones and which are which are the which are the classic ones Hi, thank you so much for that talk Actually, I've got pages and pages of notes trying to relate what you've been doing to the areas that I work in But I was thinking about particularly about your your databases where you have you know neurons And here's their properties and here's their receptors and they seem to all be Facts does anybody ever come to you and say no I want to argue with you about that point Or do you have a representation of? Probabilities like most of the time this is true or all the time. This is true We've never found a counter example or any sort of evidence like that so one of those one of those entries was pink and that means that that's Been shown not to be present That property and so that's one of the things we want to know However, there's a recent paper that we were looking at recently that in which they've looked at NAV 1 1.1 2 and 6 and those Subtypes are all present to different degrees of different Subtypes of hippocampal pyramidal neurons and so Rapidly a database can get very complicated And but that's exactly in my opinion why you have a database so that you can annotate those complications or provide models with those Complications and let people then figure out themselves how they want to let the classification evolve It's a tremendous challenge for a classification We don't want to get weighted down by thousands and thousands of different Subtypes of neurons or subtypes of channels on the other hand We want to represent what is of active interest to most people Doing research in a particular field. What's most useful to them? I'm really impressed by so many databases Perhaps we'll hear about this later in the conference, but are there are some people who build machines database crawlers using artificial intelligence to Implement Questions that you might have in mind It did avoid so you mean in silicon You mean I would assume it would be a software machine that would crawl through the database and and use some methods of artificial intelligence to Provide some sort of understanding Yeah, I don't know how I just curious how well that it is already advanced and what what's what's happening There are others here who can answer that question with far more Knowledge than I can maybe mark your part. This is Jeff Jeff here Or maybe that can be a discussion point for later on Marianne. Do you want to answer that? Here got a microphone We always we always Depend on Marianne to so in terms of employing it formally within for example NIF The answer is it has not yet been done There are things like Wolfram alpha and others that are trying to do this for larger Datasets, but you know again the idea of a NIF would be that there are different steps involved and now that we have all of this Information being sort of unified in a platform being able to build agent that would be able to go in and query it in an AI like fashion we do have ontologies and other things that That drive NIF so there are some concept-based searches and those sorts of things But I don't think there's been yet any sort of data-driven application that goes in and tries to make sense of this massive information There are some posters on Workflows and various types of analytics that we've done, but I think it's sort of ripe for somebody to do just that It's like now that we have all this data What do we do with it? Yeah, who's going to write the algorithms? No, I so I think there is a an implicit problem of who's going to put it all together and I think we can only do that by a community effort And that's why I think the IU FAR Nomenclature is is a good example of how communities come together each of those I and channel communities made their decisions and then I you far put it all together And so I think that's how how we're going to need how we are proceeding. That's how we'll need to proceed okay, I'm wondering if you think that there might be a Limit as to how far we can go in this endeavor due to our inability to get sufficient detail about About properties for example, maybe it's necessary exactly what's going on at each synapse in order to construct us The actual circuit that is performing some function We must be very careful not to drown in detail detail The devil is always in the details. However What's important? I think is to be able to work at a more conceptual level at each level so that you can Provide input to the next level and that's for example The idea of our canonical representation of the neuron That's the idea of Martin and Douglas with their canonical representation of circuits is that it gives you a essentially a working hypothesis For what is essential about that particular level so that you can then use it to to drive the next experiments Sir I have a slightly basic question which goes to ask And maybe this borrows from ideas of systems biology that we seem to have a lot of like structural models and structural Databases of the various neuron types and the kind of like receptors present on these neurons But what about behavioral models and you know trying to go from structure to behavior to function? Are there you know has there been a curation of these behavioral models? May that be in silico or in vivo and trying to correlate from structure to function why of course behavior and You know higher level higher systems levels. Yeah, that's a that's a excellent question a very Disturbing one in a way because as you say we have been focusing on structures that we can all agree on are there our entities but the even Characterizing the functional properties of neurons and systems is something that a step that we are not yet able to take because function is so nuanced in in the Analysis of neurons and systems that we haven't yet developed an efficient way to Represent that and the same goes with behavior because behavior can be represented by bar graphs by Videos by in so many different ways that I agree that that is a really big challenge for the future sorry I'm always struck by The huge amount of data we collect and yet one of the goals of science is abstraction We all recognize for instance that f equals ma is wrong But an incredibly useful abstraction one of the things I worry about about getting caught up in This neuroinformatics approach is forgetting that one large role is to get some abstraction that you can carry around and Simplify your understanding of a system rather than complicate it with all the details If I have to carry around all of the details to understand the system Then am I any better off than I was when I had the system in the first place? And so I think that comes back to the one of the main themes that I was Encouraging us to think about and that is now that we've developed many of the tools we need for archiving the details That we need now to do the integration and the integration has to focus on the essential details that then give rise to the concepts that Move us forward in a conceptual way. I agree. I agree absolutely All right one very last question And it's a comment So At these meetings we often get questions like When are we going to abstract? I think this kind of work is always dialectical in the sense that we need to go to the details