 Just to sort of give you a sort of an outline of what I'll be talking about So I had decided that I would start with a sort of few quotes and comments and I put updated here Because of what I've heard in the last few days. I wanted to sort of throw in a few comments there From there I'll move into giving a little preamble and background because you know people come from different backgrounds Before I sort of present how a simple network model setup and summary that sort of given essence of how Theta rhythms are generated in the hippocampus and then linking that to sort of more detailed network models and some closing comments Okay, so it's just an update. I've really been enjoying several of the talks and in particular Carol Goebbels comments about you know personal burden versus public good and you know the challenges associated it and the four points of Love, fame, money and nudge in terms of trying to change things sort of Resonated with me because I remember many years ago talking to actually Sean Hill in terms of collaboration And you know how we can sort of move forward in that direction and he mentioned about well I mean it's really hard, you know to sort of change human behavior But at least you can start with tools and get people sort of on board and sort of try to use things Together and that's great So it's really been rewarding and exciting seeing all the developments in the field and also facing up to the challenges in terms of Sort of what this encompasses in terms of human interactions So I really appreciated hearing about that and then also a very interesting hearing about sort of the Programs and the training ideas in terms of neuroinformatics competition neuroscience aspects and really the heterogeneity associated with it And so of course is only 24 hours in a day So, you know as much as we'd like to do everything we of course have to choose our foresight and in my case Oops, is that the right thing? Yeah So just just in an overview in terms of where I see myself is sort of trying to sort of develop bridges And so what I feel very strongly about Coming out of high school I actually wanted to be a veterinarian and I ended up going into mathematics because I've always loved math and sort of coming back and now Being in sort of in in between fields and I don't really know what I am except I'm in neuroscience And so this is what I feel really strongly about is sort of this cycling this collaborative cycling Towards conversions and translations so in terms of the models So we focus on biophysical and microcircuit type models detailed models and trying to sort of course interface the experimental data As we all do but of course as challenges associated that but really it's about the cycling that has to happen and so To me that's what it's all about and move into my first quote Which several years ago? I was invited to be part of the elife Board I was very happy about that because I think this is one of the big challenges in terms of what we have to overcome And this is a quote from one of Eve's many interesting essays if you guys haven't read this This is all the one from 2006 that I had I have pinned up on my board in my office because I think this is really important Even presently of course is that the Where you publish matters more about what you're actually trying to see and this is a terrible message Especially for young scientists I feel very strong about this not only for my own students because now I have a doctor who started a PhD and sort of you know Facing these challenges and you have all that sort of passion and excitement of people doing science You don't want it to get crushed for reasons that Not what this was to be so elife is a fantastic enterprise They're very in my opinion of dealing with being a reviewing editor for a couple of years now and seeing sort of the Openness the discussion the evolving, you know, of course It's a bunch of you know, it's not nothing's perfect right, but it's really trying to change the culture a peer review and This is so important because you know, this is essentially what sort of pervades everything in terms of human interaction And so the second quote I wanted to take was from this excellent perspective that I read not too long ago It just came out this year the scientific case for brain simulation So it's a really nice article if I'm sure you guys have all read it But it really lays down very nicely sort of the different simulations model simulators and all the rest of it but one of the things that it says one of the quotes from it that I've taken is making the point that the mechanistic modeling is still in its infancy and and and that allows me to sort of make this point about this clear and growing overlap if you like between competition neuroscience and neuroinformatics of course how exactly people want to define Computational neurophatics very very but you know by and large computation your sciences Maybe more about focus on the development of the mathematical models that you want to apply and use to try to understand the function of the system And all the rest of it where's the new informatics has made more focus on the tools and algorithms that one is developing But clearly when we develop these models you need the data or while you developing after developing at different stages That comes from that is clearly has to come from very nice analysis of in of people Developing those tools in the neuroinformatics field and of course how you analyze the data is influenced by how the models might be telling you How the system is working? So there's an obvious overlap. So that quote Takes me to the next quote that I I really enjoyed reading even though some years ago And I think it was said in one of the earlier talks to about sort of you know, this is a piecemeal approach It's not clear how we're going to get there in terms of Understanding, you know brain workings and you know how it's going to help us with neurological disease and the rest and in this Commentary by and Churchill and Larry and Larry said he was specifically responsible for this last statement Is that global understanding when it comes will likely take the form of highly adverse panels loosely stitched together in a Patrick Kilt? And I just I just sort of love that because that to me is what it's about We don't know what is the best approach. We don't know what is the best mile? We don't know I mean we just have to try and we have to try together and sort of patch it together So the sort of openness and clarity which is what I NCF is all about is I feel very strongly about But exactly how we do that and get about it is, you know work in progress so The other quote is from an editorial that was part of in the earlier stages About theory models and biology and this was a editorial in eLife that was nice to be a part of and one of the statements that Was made in here is that despite the rich history with the theory biology and models? Is that they seem to be a divide that sort of persists and you know I'm talking specifically what neuroscience and many reasons for that But I think one of the aspects that can certainly be improved is really being clear And that's why sort of put that in the title. It's for many different reasons But you know when we build models, right? We're doing a model of the hippocamp So I just say one micro circuit but it's built for different reasons the goal of the model is not always the same And that's fine, right? I mean we want that but that should be clearly stated so that people are looking at these different models Realize that when you build on it what the rational or the challenges what the assumptions that went into how the parameters are chosen What were the justifications and all the rest of it? So those things even though they are there To a certain extent in some papers. It really needs to be upfront much more because then it's hard to build this patchwork quilt Okay, so The next quote is from Eve again and even though this is a couple years old I think it's I think points to something that Feel strong about also and this gets into you know, this collaborative view is that Understanding brains and so when the era of big data clearly we have been for some time In order to get there we have to of course collaborate and work together And so she talks in this essay about being very exciting times because of all the sophisticated tools and methods that we have but also trepidation because of you know someone using a tool as a black box and you know garbage in garbage out for example So it's really hard because there's this sophistication at all levels of experiment mathematical modeling computation, you know everything so signal analysis we you know heard the talk this morning was like Fantastic, but you know, I certainly don't understand all the mathematical details and in the tensor even I know a little bit of sport tensors So really paradoxically the future is about developing stronger quantitative understanding and experimental community As well as better biological intuition of theory community It just makes it easier for people to come together and so this quote is from a while ago But I this quote really struck me back in 2000 when I read this in the Nature Neuroscience supplement Is that because science starts with human interaction? So if we want to sort of interaction collaboration and this is something we you know we heard about I mean That's why Carol Goebbels talk really resonated with me certain things she was saying is that we have to Leave our prejudices at the door and we have to and this is where sociological forces have to be tamed And so so that quote from Jean Laurent in that article is sort of I end with my own quote Or something I wrote because that is what sort of I was invited to write this some years ago But really to have the synergy we really have to have more regular discussions and interactions But at early stages not sort of oh, I have some extra data could you mod it? Oh, I have you know idea. Could you do this experiment? We sort of be talking at earlier stages Between individual theorists individual experts individual everybody But this takes a lot of time and of course open-mindedness because we have to sort of Recognize each other's hard work and that's hard to do. I mean if you don't code like if you're an experimentalist You know, it's not that straightforward to throw together Matlab code because there's certain things associated with it And similarly if you're not doing the experiments, you know to realize how hard it was actually get that recording I mean these sort of things really have to sort of be appreciated to sort of work together and Asking lots of critical questions with mutual respect. You should ask question You know anything. I don't I mean I didn't know anything about you know Hippocampal detail into neurons many years ago, but you know, I keep asking questions stupid questions But if you don't ask you don't know and so Similarly the experiment should be asking the models those critical questions on why the heck are you using this kind of models? It's not good, you know, like these are the things that should be upfront early on so that's the end of my quotes and so And comments anybody want to throw things on me or it's a stuff But I mean It's really about the human interactions right take home message And you know in terms of individual people and being open-minded critical question listening to each other And so in terms of how the brain works, I think it's clear to say You know what so what kind of neural code and it's clear that neural oscillations have you know something to do with this And we've known since back in 1929 when Hans measure measured Oscillations rhythmic activity at the brain and of course we've known that Bob Marley told us all about rhythm I grew up in the West Indies and so this is the kind of music I love and Once we talk about neuron oscillations We have to talk about inhibitory cells inhibitory networks and interneurons because this is sort of what seemed to be the controllers of the generations of these Rhythms and this is just one particular inhibitory cell the Orion's lacuna molecular interneuron in the hippocampus CA1 These are just examples of papers are some more recent than others pointing out brain oscillations and function in terms of you know I think it's fair to say they're probably not an epiphenomenon because we see sort of this high gamma frequency oscillation with working memory theta enhancement and and retrieval in phase specific ways Temporal open neocortical interactions with memory and ripple oscillations and theta oscillations in particular As probably most people know Has been studied in the hippocampus for a long time and they're very probably the most robust rhythm there And we have this very clear phase Code in terms of the phase procession of the spiking as the animal moves through space And so I think it's pretty fair to think about Phase code it's pretty clear that there's some kind of phase coding where the oscillatory cycle is like a functional unit and this has been said by many authors and Indeed it was back in 2002 that my castleman colleague sort of you know came up with this suggestion about encoding and retrieval in the peak in the trough and in the previous paper that I just cited This is exactly what they showed is that if you stimulate at different phases in the cycle to get an enhancement of the retrieval and encoding So all of this is to sort of tell you that you should care about theta oscillations. They're really important and A lot of people have studied it and are studying it continually and we also know that so that's in rodents that I was talking but it's also clear that we have them in in humans and The difference is it's even though and they seem to be functional similarities associated with the theta rhythm But except they seem to occur and intermittent about so this is something to keep in mind But the the idea is how are these theta rhythms generated? And so they're being clear an open part Then I put my title is because first of all when you say theta rhythms a theta rhythm is a theta It's not a homogeneous rhythm. There's many different types of theta rhythm So in this very old and I apologize for people who probably here who would know more what theta rhythms and myself that 2002 Puzaki one of this reviews on theta oscillations just showing that you know when you have so this is in the intact brain Recording and you see the sinks and sources giving you this sort of this theta rhythm And if you sort of lesion the entorhinal cortex, you still have this ongoing theta rhythm But of course, it's now not not coming from the region where the entorhinal cortex comes into the Campus, but you still have this ongoing rhythm So there's sort of clearly two different ways in which the theta rhythms are sort of generated or they come about And then this is just to show you when you sort of stick an electrode down Into the this in this case the mouse that you see there's different features associated this ongoing theta rhythm You have the gamma on top so you have these theta gamma rhythms But you have polarity differences depending on where you're recording from and features and all the rest of it The theta rhythms also are they come in type one type two that associated with Social or fearful stimuli so ventral dorsal differences And so there's a lot of people studying the data rhythms of studying hippocampus And there's a lot of information known and clearly we have to sort of be very clear But when we talk about theta rhythms what we're referring to and of course that means, you know reading a lot and interacting with people What we also know is there's an intra hippocampal theta rhythm And so this was from 2009 from Sylvan Williams lab that showed that you can generate a theta rhythm in an in vitro all hippocampus preparation and so This is shown here what it looks like in a behaving animal and when they sort of take it out and stick it in a dish so there's a lot of details associated with it, it's an atropine insensitive one and That's another sort of difference between the two. So what suffices to say is Even though there's many different ways to send a generate theta We do not have a sense of how it's actually where is this coming from and so this is where so the models could be very helpful Of course once you think about this. So this is a local circuit is Interneurons inhibitory cells so as I think probably everybody knows there's a very heterogeneous Group of cell types inhibitory cells We can't just sort of think of them as inhibitor and excitary cells is inhibitory cells of many different types And in this now pretty old review This is schematized showing that the three different types of inhibitory cells basket cells oil cells and axon or chandelier cells Sphere in space specific ways in this one an ongoing theta rhythm and then this is a sharp wave ripple This is the promedal cell. So there's again seems to be sort of controlling in different ways of different inhibitory cell types And of course the models have been built of them. Of course, there's a lot of detail in terms of morphology connectivity differences what they stand for and the intrinsic properties and In this paper that just came out this year. We have now 49 fine clusters of different inhibitor cell types. So Lots of excitement. This is coming from the single cell transcript So it's very exciting times because we can really get a handle on it is for inhibitory cell types But of course this week is more challenging to sort of try to understand how this sort of complex interaction gives rise to these Theta rhythms and I like this quote from Buzaki because I was back in 2006 where you know We really can't just you know when we think of these models just think of inhibitory cells like secretary cells We really have to think about, you know, what inhibitory cell type we're talking about But of course that really raises challenges. So my mantra for many years now has been neither ignore the details nor be consumed by them So really when we're thinking about these brain networks, you know The context and the function that you're specifically Thinking about the size the network size and its particular architecture connectivity and cellular characteristics They're all going to contribute in sort of different ways depending on the specifics And so how they were generated. So let me just at this point put up that knowledge month So in case I run out of time, I don't have to rush through The you know the people who are responsible to work in different details And so I like to acknowledge all present and past lab members and collaborators because it's a lot of discussion that goes on and You know different ideas and details, but for what I will show you specifically Alexandra KT Scott Anton and Melissa. I'm showing parts of what work they've done in a sort of summarized way But I was asked me lots of questions and from the collaborative perspective. So Sylvan Williams lab, of course So Kerry Hugh was a PhD student in the lab at the time. We did recordings Mathematicians Sue Ann and Wilton and now Jeremy and present in terms of sort of Discussions and I'm not showing all the details of the work But a lot of that theoretical mean fell analysis sort of allowed us to sort of come to terms of where we're situating the model And of course funding sources and support in the Institute Okay, so basically because of this whole hippocampus preparation where you sort of have this spontaneous rhythm Suffices to say I'm sorry with the assumption that this is sort of meaningful There's a something that can be discussed of course, but we have this ongoing oscillation So that means it's an opportunity at least the way I saw it was an opportunity to really build these microcirc models To have this kind of model experimental interaction to really sort of see if we can get the models to capture In a ongoing interactive way what was going on and how the theta rhythms are generated because I don't have to tell this audience, you know this sketch which is sort of based off the classic shepherd Sort of multi-scale multi-dimensional the ways you have to think about the nervous system And so if you want to think about these different levels You have to think about your models in that context and what assumptions you're meeting Making so we're in a cellular-based microcircuit Level and so what we did so this part is published I'll go through it relatively fast, but I'm important for me to sort of tell you about it So the later part which is unpublished you sort of hopefully get appreciate And if not just ask me questions So because of this situation we use this strategy where we took advantage of this experimental context of this intrinsic ongoing theta CA1 rhythm From so don Williams lab in this whole hippocampus prep We developed a mathematical models in that context and leverage sort of theoretical insights associate how they're producing their sort of Coherent output and it experiment extensive parameter variation analysis, which is of course lots of computation and so This allow this kind of interaction. There's reasons. These are dotted or dash Which I won't get into right now unless somebody wants to ask me But it's sort of this ongoing You know reasons why you're making certain choices for certain models and certain parameters and I'm happy to talk about this ad nauseam But I won't right now So suffice is to say that models we developed That that Katie was responsible for developing had to do with PV so fast-firing inhibitory cell types and networks of them Paramidal cells and excitary cells and networks of them and then sort of putting them together What I'll show you later But essentially in a nutshell the equations the cellular equations that we use were of a is a key fish type So second-order and discontinuous type and there was a reason for going that way too as opposed to starting with a biophysical type model Which again I could talk about ad nauseam We didn't start off knowing this but it was sort of the sort of ongoing interaction with Sylvan Williams lab and you know That's going there and talking and what exactly we could expect from the experimental data and choices that were made Along the way, so this is about this sort of very open So Sylvan Williams lab is fantastic guy and you know very open as lots of silly questions and all the rest of it But you know that helps you make your decisions about developing these models So what we would want the models to sort of capture which is known from the experimental data And so this is from the lab Experimental recordings in this whole hippocampus prep. So you have this ongoing Theta oscillation and then this is a similar to simultaneous recording of in this case a problem cell And you see it's not firing every time and not every problem cell a box merely 20% fire So this one for example is not firing, but it's not that it's sick or dead because you know you I Proporize it it can I can spike so you have this sparse form of the problem cells So this is advantage of course an in vitro preparation, right? You can get these simultaneous recordings of our ongoing activity and well as sort of different cell types And so this is of course what they took advantage of of course you could do Pharmacological manipulations more easily than in vivo and the rest of it But the other information that we have from this situation is the ratio of Excited inhibitory currents on the prom the cells and the PV positive cells as well as a somatosatin positive cells So this is all information that is important to sort of figure out where that balance is From the model perspective. So as I said this this was published ready But to make the point of neither ignoring the details nor be consumed by them is we decided that we would stick with just two cell types Although PV positive cells may be encompassed by stratified cells axarctonic cells And basket cells so but this is sort of what the expanse of data They were sort of targeting PV positive cells because that's what they were able to do and so Because we knew from their work of Benedict that the it was a PV positive cells that were important in sort of having this rhythm in the first place because if you silence them you Would lose a rhythm whereas that wasn't true for the somatosatin. So this is just to quickly show that so What they had is that here's the sort of ongoing rhythm if you silence a somatosatin cells here optogenetically You basically don't lose it was with the PV cells. You would lose it or basically, you know Shut it down a lot more So there's other details in here but from this we felt comfortable to start off with sort of simple network model just had the two and as I mentioned was a Isakievish style model that we chose to use because of the fitting of the parameters were based directly on recordings from this preparation Of course blocking synaptic Inputs to get the ABD different parameters that in Isakievish model and this is just showing an example of of What it looks like and in the network? the size of the network 510,000 and the particular connectivities again what was sort of known in the literature and I'll get into the size a little bit more when I get to the detailed model, but also we had other input sort of noisy input coming into the sort of system sort of Representing maybe afference from entorhinal cortex E3 because when you cut it or when you have it So like did not sort of contributing anything So As I said, this is this is out there So basically to say that when we have just the excitatory network problem alone It is literally impossible to get the sparse firing So but when you put in sort of the PV cells you can get the sparse firing and maintain this population rhythm So there's balance of where we're sitting to get get theta frequency population rhythm sort of came out of Mean-fill analysis that was done in terms of the whole system so we can anywhere we should situate things Again, that's published. I'm not going to get into that and this is just sort of the roster plot This is the 10,000 pounder cells at the top and the 500 PV cells at the bottom and in a summary of The take-home message the network size and accepting inhibitory cellular models are based on that preparation in terms of you know What's you know what they're exactly doing and then we took advantage of Theoretical what is known in terms of producing coherent output So for example the PV network is sort of the in mechanism so it like Paul for example is expert on those kind of analysis in terms of producing This sort of current output was there So that's situated where those promptive values needed to be in the promissal network on its own had some 5-second adaptation To get sort of a theta burst type frequency out and so basically and then we will get the state of rhythm in this Sparse sparse firing of the excitatory cells suggesting that maybe we're capturing some mechanistic essence of this data rhythm So therefore Because of what we have in the model spike in adaptation posting to rebound and these large minimally connected network So this is a see a one is minimal less than 1% connection between the prompter cells But in order to match the external to data. Yes, we will get that but it was also Required that you had to have a larger PV to prompter cell rather than a prompter PV cell Connectivity otherwise you wouldn't actually match the property excited or inhibitory current balances that we're in experimental data So it's important because I'll come back to that when we sort of make this sort of link So, okay, we have that so as I mentioned that's published already But we wanted to sort of get a better handle on well Is this really sort of the robust mechanism that could be going on in the model that hopefully is representing the biological situation? so together with And Jeremy had a student Anton who created this e-cell model excitatory prompter cell model database And so he did that so we could sort of create Many e-cell models. So there's 10,000 I mentioned but preserve the spike metrics of that wish that where you have the rhythm And so basically we just changed those Parameters created a large database and this is simply to show that so long as you preserve the metric And we just took very simple ways of preserving a spike in adaptation posting rebound real based metrics of where it's sitting like in terms of assessing it in the different models and suffices to say is that you You know you still have it so this is just say when all the excitatory cells are sort of the same But they're different of course is getting this noisy input But now they're all different here But they have the same metrics and you still have this sort of rate of course is not going to be identical But what we found in this process that it was the adaptation part was less critical really It was about the post-inhibitor rebound and where that really so that balance critical balance was and so this is now just sort of Four examples of just the excitatory output, of course the inhibitory Cell is there to say that where is this third frequency bus coming from in this EI network? It's essentially emerging because of the net input to the prom the cell population So long as you have this balance excitatory inhibitor input and so You at the way it is situated came from those previous models So I've to get into details associated with that But we situated it close to sort of the edge so that we knew would have current output in the I cell And so that allowed that made sense to allow us to look at this from a phase response curve analysis To really get a handle on maybe how these rhythms are coming out. What's controlling it? So what I can safely say now based on Swap doing many phase response curves is that the inhibitory input boluses tuning So long as you have the balanced excitatory inhibitory input You can sort of tune it to get the state of population output and this is just showing example So the top here is just showing you two thousand of the ten thousand and you can see the frequencies faster here So this is a network and the bottom here is the distribution from all those A B The networks that we're having there and now what I'm showing here is phase response curves And probably everybody knows what phase response curve But basically negative here means a delay and the positive is in advance for the inhibitory bolus coming in And so this is just showing for this excitatory cell versus this excitatory cell which is taken from values in these distributions You have a much larger Amplitude of the delay so that what you think you'd slow down the excitatory cell and yes you do but you're sort of advancing and Slowing it down here But if you look at the net frequency because of the different intrinsic property of the excitatory cell here It's a 8.6 and here it's 40 now I'm choosing these I values to be appropriate to where they're actually getting in the system and so basically There's a frequency that the excitatory cells can fire at for the net input It's getting and the inhibitory bolus sort of timing and I'm tuning it So this is what that is showing and to sort of bring about that sort of balance aspect This is just a another ask another example here So here it's slower, but you can see that it's kind of losing it So you kind of need to have this this balance. So again, if you look at the PRC's The phase of Sponsorship, so the hairy this is sort of slowing thing down a bit But the frequencies are not as different as the previous slide that I showed you and this is actually a slower rhythm Network, but it's almost been lost. So in other words, this balance of the EI is sort of close to being lost by virtue of The different parameter values in the E cell network, which is now required to drive the I cells So all of this I think makes me comfortable and I mean, I'm just showing you a snapshot of course that the the mechanism sort of the generation mechanism in these simple models which are Hopefully presumably representative of what's going on into the intact preparation is emerging because of this net input to the prom The cell population with its particular entrance a characteristics as captured by the ABD kilo in the is a kivish model With this balance excitor inhibitory coupling in order to have it in that state So That's nice. It's a simple model. So I feel sort of very excited about that But it's hard to sort of feedback directly to sort of experimental. What can you have the model at that level? So yes, it's representing PV fast-firing cells We know and it was based on those but of course there's 59 different inhibitory cells right in the CA1 And so it's not about trying to put them all in at the start But what was really nice was there was a another model of a full-scale CA1 circuit That was developed in event shows this lab and this was a heroic toward a forced study in many ways. It was based off of Assessing all it was known in the literature from their own experiments and other experiments of CA1 circuits cellular properties connection properties Synaptic properties, you know many different ones and it's all gathered at the table Everybody anybody can go and access it. So that's that was there in this 2013 publication and then Marianne and colleagues have put it all together in this model and Got data rhythms and what Yvonne says is that because it only okay So it has eight different types of inhibitory cells and the prom cell population 300,000 plus so it's representing the actual size of the CA1 Microcircuit and probably some people know that there's a more detailed one that will hopefully come online soon from Human Brain project and It produces data rhythms So as Yvonne said like he said wow this they just produced that kind of stuff now Of course is all the details complexity going on in the set of differential equations There's a multi-compartment models, you know, they have sodium potassium all the details associated that they're so relatively simplified to You know so the prom the cell in this case is very detailed But some of the inhibitory cells even though the multi-compartment is still not super detailed but it produces this this data rhythm and Because it's actually just a CA1 this is sort of loose I would I call it loosely based on the whole epicampus prep because again, you have this intro hippocampal theta rhythm So this in some sense I thought was an opportunity because now we could think of this sort of overlap between the detailed and simple models Okay, all that both have advantages and disadvantages in different ways. So in order to do this So what I I maybe didn't say directly is when we did our simple models when Katie had put this model together Some years ago based on many aspects We made the choice so we're thinking about size and so we intentionally created a One cubic millimeter is what the estimate we did and we felt justified to do that because of the experimental work That said if you use proclaim and you block in certain parts that you can you would still get the oscillations In other words, you don't need the entire HIPAA campus is just a piece there So we sort of focus on that so we could think about cell numbers in a particular way and actually that's about 30,000 but as you would have seen those 10,000 but the 10,000 we took advantage of because invariance in the mean field analysis model So basically in order to now make this comparison, we have to take a chunk of the detailed model So the detail model is 300,000 so I'm gonna take a chunk of it So this is a very large-scale model high-dimensional and so you just take a chunk It's not like you just just get it right so in terms of trying to sort of understand What's going on the first thing is you know, is it valid to sort of even think about this? So one of the first things we did and this was Melissa Was to compare the connectivities So in our model, we just had random connectivities, you know 1% or less than 1% for this connection 12% was sort of based on what's reasonable known known about the connectivity and then we had Different values there that we explored In the bizarre model there these empirical fit empirical numbers that were based on their Intense work in terms of knowing what was connected to what where the dent rights and all the rest of it Okay, good because I'm close to being that Five minutes of talking or five minutes of yeah, okay, so that's no problem So basically in a long story short what Melissa did was she sort of went into the detailed model and again Everything's open and accessible so it's important to you know be able to do that I mean this that's some of your technical issues But basically from that she was able to extract and what is really nice is that what we had found in the simple model That I had mentioned before was that in order to be consistent with experimental work You had to have the PV departmental rather than the problem to PV larger That's exactly what the empirical was there so sort of felt good It was consistent made sense that we can make this comparison so we went ahead and did that So in a nutshell what we did was we took the guidance from the simple model So we did a comparison of the synoptic weights which led us to focus more on excitatory connections And so Alexandra, you know once you kind of got this working She was able to generate these theta rhythms in this chunk very nicely, but in different ways And I think that's what's important. So a couple of things one The fact that you have a smaller network so 30,000 rather than 300,000 means you can do many many simulations Which is what you want to do to be able to sort of explore the palm to space So that's what of course they didn't do but because it was huge right so there they did a few perturbations So in units exploration in the particular way of focusing on the excitatory connection So the x-axis here is the problem the problem the connections and the y-axis here is the Inputs sort of if you like the noisy input enter on a cortex da3 into the problem cell population That's what that's representing that's what she's exploring because that's what the simple model said of where the focus needed to be in terms of getting this sort of Theta rhythm that we had here So this is the north the theta power normalized to power Frequency and amount of that stimulation that needed the stimulation I need to come into me to get that theta rhythm And so this is circled here, and she calls it high and low just to show you get this theta rhythm But it's in two different ways, and this is nicely shown when you sort of start blocking things So this is just now showing the low So spectrogram so this is the raw LFP output. This is the theta filtered and the theta gamma filtered and here's a See the oscillation there when you do sort of the FFT If you now remove the PV cells as I mentioned the PV cells was important We knew from the X-ray to bring bring this about you see that you basically get a huge reduction in sort of the slow A theta rhythm and actually get this a larger gamma rhythm there And this is sort of what the raw LFP looks like in this case when you block it So so this is so you know there's many details of what you could look at in different cell output This is to make the point that the low and the high so to both of them producing theta rhythm But when you block the PV you get a very different response in terms of what kind of rhythms comes out So in other words a mechanism of sort of the different pathways a cell specific path There's a producing these rhythms are very different Okay, and so in terms of what's happening in the biological system This now gives us sort of a way to sort of tackle that more directly a cell specific way even though It has eight types of inhibitory cells, but again, hopefully that's sort of enough to sort of capture The biological complexity and then this is now just showing if you get rid of the problem the problem connections You sort of lose it So I'm almost on my last slide right close Oh So basically this is I put this up here to remind myself to say that it's a simple model guidance The the detail models of how to export and what we could sort of get away with and I and I to me It was very exciting when Alexander sort of was able to produce this in different ways And that you can see these differences from the different inhibitory cell types Because that's now what we could sort of more directly map went to the biological Preparation and hopefully move forward from there. So my closing comment is I actually didn't put this but this is from Schematic that I made for a 2012 Little little review that I wrote and you know really about trying to have these two-way arrows everywhere And this gets back into the whole human interaction part that I talked about but really it's always a balance, right? So we started off with these simple models, but not because we want to do simple models as reasons Associated experiments of data when he developed the model in terms of what we'd sense what we could get out and what we could see And then limitations computational wise if you want to do hundreds and thousands of simulations as largely turn that to really explore The problem states and see where things on how to how to explore it And this is coming from the theoretical analysis theory and the analysis to look at it And so this could only come to being I mean we're not there yet But this is what where where I like to try to be is without having sort of clarity about the modeling experimental data The computation and being open and clear and not asking lots of critical questions of each other And so that's the end for now when I really have Okay, thank you for answers Yes, I'm sorry if the question is not directly on your talk. It's actually not my expertise in this area of But I was I mean always a bit wondering what what are the the standards and the tools that are you think are missing in for you to be productive and you to Assemble and and and link all those those results and is there and What would be the the thing that you would hope to the field would be developed such that you'd be much more productive in That kind of line of work. Mm-hmm. Well, for example an electrophysiological data If it could be more standardized and I know of course is NWB Efforts in that way But all the metadata that goes with any of these excerpt experimental situations because you know We have these direct conversations going on with collaborators and it takes a lot of time and that's fine You learn a lot in the process, right? But you know, it's a huge effort for them You know, so they're giving us a data. We're discussing. We're asking somebody to give oh, okay Oh, yeah, and the juncture potential is that we can calculate this You know, so there's a lot of ongoing conversation Which probably could be reduced if there was sort of a standardized way that people sort of you know Had the electrophysiological data with all the metadata, right associated with it But that is a huge effort right for experimentalists and and the same on the modeling side They should really you know besides it's not just having a model parameters But having the experimental data that you use to decide why you use certain parameter data parameters in your model Yeah, do you envisage a time where and if that standardization is good enough like Computers could actually analyze data directly without your Human interface well that that would be nice, but I don't think it's a matter of sort of getting rid of the human interaction I think you always have to have the human interaction at early stages, right? Because it's more sort of defining sort of the questions and how you're going to go about it, right? So I mean and I didn't talk about it today But instead of other modeling studies with the internal specific one of these Inhibitary VIP positive cell models that we developed together with Lisa to Paul next lab There's a lot of conversation in the beginning of what she was thinking should get out of it What I said what potentially possible giving what you know and there's a lot of this going back and forth So that wasn't about the experimental data was about seven initial conversations and that's why I think the earlier the conversations go on So that you're at least on the same wavelength of what it is You think you're trying to do and what you think is actually possible based on you know What I know about modeling what they know what explain what is possible what is expensive and you know everything there So it was not a matter of getting rid of it But you can certainly reduce the amount of I would say grant work that you know students for stocks You know other people have to do to sort of be able to share the data and the specifics associated with them Yeah, so I that would be wonderful Do you have another question? I just have one comment. I mean actually I liked very much your your observation about I mean it's obvious that we should talk. I mean the theoreticians and the experimentalists But very often you know even if it happens it takes part very late in the whole process I mean so we have come several times across the situation but experimental colleagues come to us and they say oh, you know We have those very nice experimental data. Can you help us understand them? I mean say yeah Yeah, we'd actually do much more, you know here to do this and this two years ago and you were starting the grant Yeah, and now of course the whole work has been done, right? Yeah. No, that's that's that's I that is so important And this like the next step is just yeah talking early and so like having these interactions. Yeah. Yeah, no, no I mean, you know I've had that happen to me with you know other people before but you know That's not going to happen unless we talk really okay, so