hello and welcome to another episode of Max plun Florida's neurotransmissions podcast I'm Joe Schumacher joined Again by Dr Jeremy Chang how's it going Jeremy pretty good we've got a really great show uh lined up for today joining us today for the first time on the podcast is Dr Tim Harris he's senior fellow of at hhmi janelia research campus and a research professor in the department of biomedical engineering at Johns Hopkins University Dr Harris thanks so much for joining us on the podcast today happy to be here so you know we love to talk technology on this podcast uh how it influences the directions people take in Neuroscience um but to sort of back up a little bit of um where you come from and what how your career has been informed um it's been very clear in a few episodes of this podcast that throughout the history of the last few decades in Neuroscience some very impactful researchers who've really charted important directions in our field all kind of originated at Bell labs in um uh in New Jersey and uh you're one of those people so I was wondering if we could talk a little bit about um earlier days in your research career when you're working at Bell Labs what was the environment like what was special about a place like Bell Labs back in in the the 1980s it was special for a reason that I think most people don't know first first of all it was a very very strongly Talent filtered pool yeah so you did not have to worry about being you know the smartest person on your hall or the smartest person in your building you needed to worry about were you the smartest person in a private office um but I mean so that sets one tone um but the thing that made it special for me was the very wide variety of of science that was going on and that much of the invention of that science was done by the people that were there and that the groups were really small typically one principal investigator and one assistant and groups were never bigger so um and there was a policy that was implemented in the 30s and 40s on purpose called the open door policy there's a book called The Idea Factory that documents this when I read that book I thought oh yeah I noticed that culture the day I arrived the rule at Bell Labs were you got to talk you don't you cannot close your office and and try and work alone and so if somebody knocks on your door and asks you got a minute to talk the answer is by Fiat yes so my favorite example of that is one day I was worrying about a really Arcane thought about what happened when light was uh shined at really small metal particles and I thought if they were super conducting Metals it would be just really awesome so there was a guy in my department that had been at bellab since the late 40s and I went down the hall and asked him Darwin would I said Dar do you know anybody who knows about the optical properties of superconducting metals and he said oh yeah Jim Allen invented that I thought great Jim's like down the hall and around the corner down the hall around the corner bang on the door the doors open you got a in it yes I do I got this cont obligated to have a in it yeah but it was it was it was part so no one ever said you can't close your office door it was in the water like you you can't close your office door you got to talk and so and so I said Jim I got this great idea there's this funny phenomenon it works when small metal particles are illuminated with light you get these big effects you know it seems like it would just be really big in superc conducting metals and he says oh they're just ordinary Metals I said really like but they're superc conducted he says yeah I'm sorry but at Optical frequencies they're ordinary metals and I thought oh man it was such a good idea and so in 15 minutes I had a great idea I found out who invented that I found out it wasn't an important idea at all and went back to work it's a very uh emotional roller coaster to go on and well it was the the the really magic thing about it was is that you didn't have to think was was that really right like you were talking to the guy who knew as much about that as anyone alive and so you could just take it to the bank okay Jim Jim understands what this is so or any you know other you know so this particular phenomenon worked best with gold and silver and but but they don't superconduct um but superconducting Metals don't do what gold and silver does at room temperature and so so it was an example of there was only one person out of the thousand people in research that I ever encountered that only answer the very specific question you asked and wouldn't think with you and it just made it the most interesting challenging you know I went there to survive I thought well this place is like way over my head um and and so I thought well if I work really hard you know I can last 10 years before I'm burned out and then I would have had a chance to work at B abbs so one of my friends said one time lucky for you and me Tim science is not an IQ test and I thought well lucky for me Alex I think you'd do fine [Laughter] but I think that science is not an IQ test I mean there are people who you think my gosh you know where was I standing when they were passing that kind of talent out but science is an importance thought experiment like what matters there are infinite questions only some sub that are worth working hard at and and so that's the that's the issue that I think is what question that's important is answerable in the near term there's no point in trying to answer a question you don't have any access to you know like let's try and go to Pluto thinking well you know we can work on it we're get anywhere yeah exactly the example that I like best is that let's assume you're a pharaoh of Egypt I invented this when I got to janelia because Neuroscience is so hard I thought we should just quit and wait a hundred years when we're smarter I got a lot of push back on that um so I said okay how about this let's say you're the pharaoh of Egypt it's 2500 BC you got a lot of extra wheat nothing to do what shall we do well they knew about Mars they knew it was a planet uh and and they could say Wow Let's try and go to Mars and and somebody else would say I think we should like build a really big stack of rock shaped like a cone well if they have tried to go to Mars it would have been a pointless task they wouldn't have made any progress but they made some really great piles of rocks because it was it was feasible with the resources they had which was Labor and rocks and spare time outside of the farming season and so so I think you got to think about what's doable what questions can you get some traction on and then you know work hard and and be willing to be wrong think all right you know what we thought we knew we don't we're just not making any Headway you know what's Plan B so when you're thinking about what's doable you know I think a lot of what's really great about you know innovated technology is that it changes the landscape of what's doable is part of your task to focus on an area of research that is doable in the short term do you also think the next step ahead and say what doors does this open what questions does this make available to us later or are you more focused on the problem that's right in front of you I mean do you are you conscious of technology is and it's potential to be transformative in our way of thinking or are you more focused on you know um a research program because it is doable and it's an interesting question in and of itself or even I mean Bell Labs was a commercial uh commercial entity right were there any sort of considerations for like you had to make an argument that this was going to enhance the Communications business so I would to work at Bell labs when it was Mell AT&T I was one of 1.3 million ion AT&T employees um the biggest company in the world at the time I think that foxcom is now approximately that big but um you had to make an argument however thin this is advancing the the the the telephonic communications capability so a bunch of Material Science some of it was pretty thin but nonetheless I Was A Material Science organization we were trying to make stuff that could be used in a Communications application uh you know the transistor was invented there absolutely on purpose because they needed a switch and the switches they had were vacuum tubes and vacuum tubes fail and so you needed a switch that didn't fail and so it wasn't like piie in the sky physics like we need a switch and like you know John Bine would calculate what a switch might be and then they'd go on the lab and try and make it and the next week they said well we we assume this that doesn't appear to be true we assume that it does appear to be true where do we go in our next iteration and so when I think about this I think my friends want to record as many neurons as possible they do not have a technology that allows them to record nearly as many as they would like so what are the available Technologies what do we know and what are the unknowns that we can resolve to understand is there a path forward and so the way I think about is is I you know I think about okay what are we trying to get to what are what are the unknowns to get there and how do we understand those unknowns to find out yes that's a path and we should go there or yeah that's just not as important as we had hoped and we're just not going to do it because it it's not important enough for the effort and so my usual modus sarandi is to find people who are I I'm a measurement tool Builder and so the question is what measurement do you want to do do I would ask people when I first got to janelia the hhmi investigators this Elite group of people that hhmi funds you know what do you want to measure that you can't measure half of them told me no one had ever asked them that question I so this is when you started at jally early on what do you want to measure that you can't measure I'll think about do I have any ideas about how to get towards that objective because I would like to invent the measurement tool that's important you know I have a list you have to vet it biologists sometimes are really bad at at picking what's important versus what their favorite hobby horse is so that's where my judgment has to come in thinking yeah I understand you want to know that but no one else does so I'm not going to spend that's sort of my my game plan in that space you know backing up like kind of slightly in your career so you know telecommunication and then um you know uh positions looking into um you know biotechnology largely after that what was the motivation with J with janil like this was your kind of actually your first for into Neuroscience specifically yeah so how did you get the lesson in this is first of all that don't worry about not knowing anything about what you're proposing to do you can learn it that's very inspiring so no so I when I got to so so I started out as an analytical chemist I went to Bell Labs if you do measurements for a living and you're at Bell Labs you're going to do semiconductor devices and materials so I did that for a while I became a semiconductor physicist after a fashion many people assume I'm a physicist by training um one of my friends Lewis Bruce was had had just in invented something that are now called Quantum dots they were pretty crummy at the time he came to my lab and said you know I have these I'd like like to try and understand them your equipment seems to be all right it's a classic Babs thing you know that you always go look for help because you don't have a big group and so I thought well that's fun and we were struggling with the fact that in Quantum dots their size determines what they do and they're not molecules so they're all different sizes and so they're just a heterogeneous mess so somewhere in the late 80s I thought well do them one at a time we'll do them one at a time and then heterogen 80 doesn't matter well you can calculate the signal from this is the classic problem calculate the signal signal is big enough what's the background are we trying to do astronomy in the daytime or we trying to do astronomy at night astronomy at night's pretty easy astronomy daytime is really hard so we set out to understand if you wanted to look at a Quantum do one at a time could you get the background low enough to see one we knew what we were looking for interesting enough with the with the history of Max plank um aan n told me a very similar story about about looking at single ion channels they knew what the noise was in a multi-on channel patch signal so they knew the magnitude of the signal they were looking for the question is could you see it in the absence and so I was really stunned at how similar our thinking was well Eric bedig was was there at the time I didn't know him and so I had a new post do and he came to me with a paper that showed an exotic form of of uh Optical microscopy that would have hypothetic Ally made a low background so I looked at it and I thought man this looks really hard like let's do an easier experiment a couple of months later he discovered that the guy who wrote those papers was down the hall and around the corner so I said let's go see him like immediately bang on the door like hi you don't know us we're interested in this he was trying to build a transmission microscope that scanned really fast I said nobody cares like that isn't in fact turned out to be true but I said I'll sit in one place all day but I want to do One MH he in my postto invented a new technology 10,000 times better than the one he was had in hand that is the technology that allowed him to image single Quantum dots or and that was second the first was single molecules the first single molecule image ever done by Eric big got Bell ABS he was smart enough to figure out how to turn it into a microscopy project and won a Nobel Prize eventually we we did it we we used it to do single Quantum dots and discovered they blinked we didn't know they blinked until we saw them one at a time and so it's the classic case in my mind of okay you know half of the answer you know what you want to do and why you have to go find the second half of the answer can you manage the background it turned out that the background management problem was just not a big deal and we should have not gone into exotic microscopy but it all was very fun in the end Eric's a really fun colleague to be around and so no nobody suffered from the from the diversion we took but that's an example you know something you you know what you want to do and why and then you have to knock down the unknowns to ask can I get there yeah and so you know this the the sort of segue then after working on you know um you know DNA sequencing and other Tech Technologies specifically into Neuroscience what was the motivation with with janelia well threefold one is I was working at a biotech that I thought was going to crash okay good start so I thought these guys are not smart enough to turn this into a business and so we're going to crash and so I thought I don't want to be part of that crash and so I thought okay I need a new job what should I do I had left Sol State physics to life science on purpose because I thought look I'm not good at measure at making stuff I'm good at making measurements and so I want to go where measurements are the primary problem that's biology MH the stuff is already there you need you need measurements that's interesting I I feel like a lot of theoretical physicists also move into the life into Neuroscience specifically around these same time frames but for slightly different reasons a lot of people with you know chemistry and physics backgrounds find something tractable in neuroscience and for you it's actually measurement it's it's um it's not building building this theoretical framework to explain Dynamics or dynamical systems it's more an application of of a neuroscience is really hard you know I I was arguing let's give it up for a century and see if you know and so it it's a measurement limited problem in every way and so I had to make the bet that look my measurement capabilities are good enough that we can make some Headway here it's not Egyptians trying to go to Mars it's you know meaningful progress and so I knew as much Neuroscience as you learn by reading the gossip pages of Science and nature I mean that is an absolute truth you got to be willing to ask bad questions you got to be willing people think God how can you be that ignorant I'm thinking I'm sorry I am that's where we are I learned somewhere along my career to ask good questions fairly early in my literacy um possibilities um and so I I don't know how or when I gained that that i' take no credit for it it was part of it be just not being afraid to ask silly questions you got to I like to tell this story I was in high school and and something that would have been kind honor chemistry at the time we didn't use those terms I'm really old um and so we were in this class of 20ish people and the instructor said okay did everybody understand that we all sat there like lumps and he said look I know you all didn't understand that so you have two choices you can either look ignorant or be ignorant I raised my hand I'd rather look ignorant than be ignorant a really important lesson that I have remembered it doesn't matter if you look ignorant if you are because you can fix being ignorant you can fix being ignorant yeah that's my career advice fix being ignorant that's great you'll get there it's very succinct and I feel like it applies to every domain of Science in general um so you know you mention when you get to janelia you ask your colleagues what is something you wish you could measure and they're they're like I nobody's ever asked me that before um I'm assuming that along the way this translates into the early ideas for your neuropixels you know what were some of the thing feedback you were getting from colleagues at the time in terms of um you know uh limitations in the capacity of like the numbers of contacts and things like that so so it's really straight forward I mean Neuroscience is about activity and Anatomy mhm and so you know they wanted us to fix both can you teach us the anatomy and can you teach us the activity so they would tell you here we are in anatomy it's a really bad space we did some work that turned out to be not as important as we had hoped I mean people ask us make this microscope it will help us do Anatomy I was a microscope engineer until I got to janelia um and so so and so that turned out to have less impact than I thought because making and labeling the sample was harder than using the microscope and microscope worked fine it was making the sample that was hard that that's even true today even with a lot of like the super resolution techniques is like it's labeling Limited at this point we can see the smallest stuff we ever care to see but how do you specifically label the chiste of the mitochondria exactly and so and so you know just completely you know vicariously for me is that dear renberg who's now at NYU was the group leader then at janelia and Albert Lee came to me and said you know we have these things they're called silicon probes they have you know four channels um we need way more capacity I had never heard of these things I looked at the technology it was stunningly primitive 60s n in 2009 it was 60s fabrication technology almost 5050 years old and so I thought well look I'm I'm not a semiconductor engineer but I think we can do a little better than this and so I set out to make the probes that heiko uses here um it took me a few iterations to get there one of the nice things about janelia is you got enough money that if the first time it doesn't work and the second time it doesn't work you get a third dry and so and but it became clear fairly early on that as far as I was likely able to go it was still not what we really needed what we really needed was hundreds of channels not tens thousands would be better but at least hundreds and so while I was finishing up the okay I'm going to draw the line here I'm going to make a 64 Channel device because it just doesn't make sense to go further with the Channel with the capacity I have access to you know there are real microelectronics companies in the world they make stuff that is radically more complicated than I can access and so I went look I you know I do what Engineers do I wrote a specification I want a device this big it has to have this noise level has have this size you know and went around trying to find someone who could make it and I ran almost accidentally across this organization in in Belgium called iMac and they were trying to make something that was relevant to Neuroscience research and failing and so I walk in the door I mean this is another lesson in life I walk in the door with hhmi written on my forehead well look Brands matter you know like Babs like brand if you call somebody up and you're just the rookiest side is possible if you say this is Tim Harris from B lives you're probably going to get your your your call answered that's nice well if you're Tim Harris from hhmi and you walk into IMC you know you're saying look I'm trying to get someone to build a neuroscience tool for me and and the management says well we've been trying for the last five years and no one seems to care about what we build their notion is well this guy's from hhmi maybe he knows what's worth doing and so you at least get a hearing and I explained what I wanted and why and the other thing that's important in my career track about about being at biotech startups is at biotech startups your spend your job is to spend money and buy speed buy speed buy speed speed is really critical and so and so you know don't try one thing try three things because if the first one fails you don't want to have the second one waiting in line you want to have it running in parallel and so you buy speed and so I thought you know so these guys came back to me with a proposal after some iterations and said okay we can do this for you $5.5 million and that was at the time just a ridiculous ask in a neuroscience tool context but it was a perfectly reasonable number in a in a biotech startup environment so I just went down to my boss and said boss I need and half million dollars and you know I was simply willing to say that because I was convinced that it would be a transformational tool if it could be made because they were already doing something that was almost just like that it just needed to have more capacity um it wasn't I needed to invent the car I needed to invent one that went faster I mean we already had the thing we wanted to do we understood it the data digestion was not going to scale very well but it was all there it just needed to be more so I went out to look for more so how hard was it to convince the uh well I guess was there a Divergence in how hard it was convince the the neuroscientists in terms of the usefulness of this technique versus the people that would fund this so there's some really interesting Recollections in that regard so when when this was all ready and I and I told my boss look you know I Have This Confidential proposal that I need to tell the The Institute about you know 100 people came and I walked them through this is what I want to make and I need this budget to make it 100 people in the room 98 of them are neuroscientists they would say science fiction you will never get that to work two of them were for former semiconductor Engineers their question was what's hard about that just again you know this very weird broad background that I bring to the table is is that you can talk to both communities if you learn their language so I didn't know the Neuroscience language when I got to janelia but you can learn it sure and so you know you can say well I look I know you guys are skeptical but I I know enough to know the semiconductor Engineers think this is a straight forward thing it's expensive to do development it's not expensive to make once the development's done and so so I convinced the bosses look you know I think this is doable my boss told me um if if you think it's that important you should be able to find a partner I told the story at my talk yeah it's actually a great story so so I think this is again one of these perspective things is that the person in the next door office to mine at the time was cydney brener who's no longer with us Sydney was just a spectacularly good scientist he was the head of the of of a of one of the most famous Laboratories in the world at the time the MRC at Cambridge University he was the boss of Frederick Sanger until a couple of years ago the only person who had ever won two Nobel prizes in chemistry so so uh Sydney was was unintimidated by risk so we thought this has Janelle you're written all over it Tim I said well Sydney Jerry thinks I need a partner and he says I got to go to MIT tomorrow I'll be back on Friday let's talk about it so he goes to MIT comes back back he says Kristoff cook announced a big project at the Allen Institute he's moving there I asked him what are you going to do for a probe he said I need a probe he says you should call Kristoff Tim so I called Kristoff Kristoff says I think I could pay for a third I don't think I can pay for half so I thought great you know like one partner is infinitely better than zero Partners so it took nine months to get the rest of the money but but again this is an example of so we went to the UK it turns out that was a janelia guy that had moved to the welcome trust he invited me to come and Pitch to them there's a second smaller charity called the Gatsby Foundation if you've ever been to the UK you might have been to a sainsbury grocery store the Gatsby Foundation is the property of David sainsbury Lord sainsbury of tural he was the science minister in some labor government while ago and so anyway he had a neuroscience thread and so when we gave the pitch there um his Neuroscience liaison came out and said I'm in for half a million pounds I don't care what what welcome does and so this is a I think a wise life lesson so now you got hhmi you got the Allen Institute you got the Gatsby Foundation it's really tough to say you don't want to join that club like yeah they're not good enough for me it's just not going to work and so it took a while to get over the because it's kind of a bureaucratic organization the welcome trust although they tell me they're the least bureaucratic organization in the UK pretty frightening thought um but but the point was that sort of community credibility generated $5 A5 million dollar in funding and5 a half million dollars to send to iMac I mean hhmi funded me and my group to do the testing and development write software the welcome funded um you know the people at UCL and in Jon O'Keefe's lab to to do their part and you know Allen put up their money themselves besides sending a good chunk to a good chunk to to imech and so it was it was a a budget that would not have been considered feasible for NIH grants for example but nothing is rewarded like success and so we were quite conservative about the technology we knew look we're not going to get a second bite at the Apple if this flops right and so and so it was basically a cut and paste Electronics design we didn't try and make it too small we didn't try to make it too fancy we had some noise specs that we had to meet and so and so it turned out to work more than good enough and that gives you a second go okay you guys now that we have this one it works great we think we can make it a third the size or or maybe less and and give it some more Shanks so it's better for MCE and so that was a much less difficult argument and neopixel 1.0 was was a huge hit I mean people were um it was like it reminded me of when optogenetics became widely available and and and viable just something that you can pick up off the shelf and immediately see like a benefit to I think without being arrogant you know people have said look this is like optogenetics it's become the dominant technology in the space and it's because it's easy and purchasable right you just order it you plug it in you don't have to develop yourself you don't have to like follow a recipe you know you can get a tool and you can put it in the brain half an hour you can be recording neurons so in neopixel 1.0 describe the
configuration it's a silicon probe uh single shank right and it has how many contacts it has just under a thousand on on 10 millimeters of shank and at the time comp competing products on the market might have had 16 64 maybe 64 was the state-of-the-art that came from my lab MH okay um elsewhere it was 16 like commercially available commercially available and maybe you'd get two to get 32 Channel like that um so quite a Leap Forward you know most of an order of magnitude you know before sort of thinking about like what are some of the other technological as you said like you can start thinking about 2.0 you can start thinking about other more um application specific uh configurations what out the gate or in the last you know uh few years have been some of the most surprising or most like exciting uh directions people have taken the technology in your perspective well I think the the question is you know where in the brain does that happen that means that if you want to answer that question you have to record brainwide and by some definition there have been a few papers that that have made that argument they do it serially they do the experiment over and over and over and over again and they record I mean they try and make a really reproducible experiment and then they record all over the brain during that experiment and then they make the make the assertion you know this working memory happens in anterior singulate cortex because if we silence it you can't remember except for it turns out that if you silence one side the other side takes over so you got to silence both sides or it doesn't you know but on the other hand if you record without silencing anything you only see activity in one side so how the how the brain knows turn on the other side because I'm I'm I'm having trouble over here it's a little mysterious but nonetheless and so so it has been these attempts at brainwide surveys that are pretty arduous you know serially and I'm the author on on at least one of them and it's really First Rate work but it is really tedious months and months and months of data of data acquisition and and the obvious thing if you're a technologist saying look we want a short circuit that we want to do them all at once to make sure that that the synchrony that you really would like to know what's in the back of the brain happening within a few milliseconds of what's in the front of the brain you can't answer that question with a Serial recording if you can record brainwide then you get the answer to that question and so and so what I like to to tell people is I'm trying to enable brainwide recording brain wide would be you know all the neurons all the time and every you know you can't do that but the idea is at least three-dimensionally 30 identifiable brain regions simultaneously maybe more I mean that that in a at a scale that that it would have been unfair to say brain wide before that and you're you're getting that brain wide because of the the length of the the neopixel prob we just have Channel capacity now is is that in the neuropixels probe you have four Shanks so you get more tissue coverage it's got 5,000 sites instead of a thousand but you still only have 384 channels and so you get 700 microns on each of four Shanks well now if you want to do the whole depth of the brain it takes eight steps to walk that of a mouse to walk that 700 microns up from the bottom of the brain to the top of the cortex and each each shank is basically like a small tiny needle with the each shank is a needle that's 70 microns wide and 23 microns thick and it's got 1,60 1,80 SES on it there are switches under every site so you get to choose which of those 1280 you record from there are only 384 wires out of a shank so you can't put more than 384 channels on any one shank and so since it's a 384 Channel probe if you want you can put all of your channels on one shank and ignore the other three Shanks that's a a doable experiment um the the point of the of the latest probe was look we'd like to record from the whole depth at one go right and how much capacity you need for that and the answer is somewhere in the neighborhood of 600 sites on a shank so we decided all right you know 1536 is a nice round number we're going to make a 1536 Channel probe we're going to make it small enough that you can put at least two probably four on a freely moving mouse and then the but the real reason behind the the technology of the current neurop pixels that I'm working on is that these are analog circuits like the analog parts of your cell phone and analog circuits technology plateaued about 15 years ago at a specific technology and making it smaller doesn't work because it gets too noisy so these are these are really really good quality amplifiers that are in neuropixels probes and so through the argument I made to the NIH who funded this project was look we know where the last chapter of this book is we might as well just skip the intermediate stuff and just get there and then we're done and we will engineer it so that it's reusable and so that's how the current neurop pixel project got designed is because there was an obvious endo and and I don't think I know of a way to do you know another factor of three or five I mean you might do 20% but 20% is not worth $5 million of engineering R&D I think we're where we're going to get for the moment and now we have to get better at using them and then if you think about basically you know continuous contacts like if you want to think about the sci-fi version of what you could do with like electrophysiology maybe that's the time you say Let's Wait a 100 years and we'll revisit it um you know one one thing that really stuck out to me was the amount of data you're collecting in a per minute basically and so I think I think the the the ballpark figure for an neopixel 2 recording over an hour is something about two terabytes of of Rod so that's for the new one that's for the new the old ones are you know 10 many tens of gigs so one of the of the neuropixels that are about to be launched generates about a half a terabyte an hour and so we're thinking about you know four to eight probe experiments so 2 to four terabytes an hour you know you record three days a week couple hours a day it adds up so as much as you might be a hero in the Nur Neuroscience Community are you like a villain in the IT department Community like so have you seen people have to make uh new types of accommodations for the type of like data infrastructure they have to support um this type of Technology well my the the guy who built janelia Jerry Rubin complained that he had spent two to 4% of a very large budget every year from day one on storage on storage because if you think neuropixels are bad look at highspeed Imaging I mean it's really yeah yeah it's really bad so so we're not the worst offenders in the building but but nonetheless it is a a you know I used to think about well you know I know how to do lossless compression of a factor of two but like who cares but if you got 10 terabytes like five terabytes is better I mean there is a really big cost difference when you're talking 2x at those kinds of volumes we're thinking hard about can we compress the data 10x and actually retain quantitative ly valid answers the only way to do that is to try it out and ask what's the before compression answer and the after compression answer if we want to build Wireless Systems compression might be either absolutely required or extremely handy one of the two um we're worried I mean it's one of the things that are ongoing right now is how do we create a tractable data volume in order to Advantage the future are they just raw storage issues or you know wireless issues or you know 18,000 Channel rats are just going to become completely out of control and so I think there's a whole bunch of of you know field work to do in order to make this new device which is kind of plateaued against a technology barrier to make it tractably usable for everyone the very best richest Labs can always wangle a way to do it but I think that you know the the Community has put a lot of money behind me to make these devices and you want them to have the broadest Advantage possible you want to build fancy roads where only three people have cars and so and so our current project is to ask how to make this absolutely usable for a secondy year graduate student at a pretty good school that's great yeah so what exactly are the things that you're trying to uh uh build that would allow that user to to use your systems so they fall into a few Cate ories one is this you know surgery of too many probes I mean that it takes some hours to do this for one and if you're going to try and do eight the people that I know that have done eight said it took of order 15 to 18 hours that is not a one-person job if you're me um and so the people that have done it well have done it in teams of and so we want to speed it up so that it's really really reliable as good as it can be in eight hours for one person the second thing is this data digestion problem you now have so much data you cannot inspect it you have to create automated tools that you can sample it and ask it look did anything go off the rails but mostly you know you have to trust the automated outcomes and so we're building tools to do automated data digestion where it's historically been done with manual visual curation you know neuron by neuron and so we're trying to figure out how to quantify that so you can say you know this is is a 9 9.0 neuron and that's an 8.8 neuron and this is a 5.5 neuron like in terms
of single unit isolation and that sort of just you know how much you should rely on those spikes being the the full reported activity of that cell do you have ones that don't belong do you have ones that are missing yeah yeah so so coming straight out of the neurop pixel you're getting this raw signal of the voltage off off of each pad and then you're creating these tools that are allowing you to take those signals to a find individual n neur yeah so so what neuroscientists want is a Time history of the activity of every neuron in the space when did they fire because they're digital devices they go bang and so the question is when did they go bang and when did they go bang relative to the other neurons in the system and so and so in the end you want accurate Spike time histories of each neuron isolated from all its neighbors that's what I can't defend absolutely that as being the most important information but a vast majority of my Neuroscience friends tell me that's the most important information and so I have to go with it yeah that's our current assumption yeah um well that's fantastic um I think you know one thing that you emphasized is there a lot of directions you can take this technology for specific applications for instance for a very thick cortex if you wanted to you know record from across all cortical layers you need a really long probe um if you wanted to be able to have some kind of um inreal time verification that you're recording from the same type of cell that you think you are it's a a cell that's you know optogenetically tagged in a genetically specific way you can integrate um Optical stimulation along the the the fibers of the the probes of the the neurop pixel probe um what are some of the sort of dream directions you you hope to take this technology um into the future well it would I mean they can be dreams what what what people want is is I didn't put it on my list but it's all of the activity all the time identified by neuron type MH now neuron type is is a fight all by itself there's many ways to say neuron type we used to we used to have this argument at janelia early on and finally Jerry Rubin just said look you know I will take two positions I will take there there are two kinds of neuron there are two kinds of cells in the brain neurons and glea yeah and then I will take the position there are as many neuron types as there are neurons and so and so it's it's a it's a bit of a you know there are clearly important classes of neurons like paramal cells are self similar but there are different classes of paramal neurons there are inhibitory neurons and there are excitatory neurons so you would love to know that without having to do any extra work so one of the things we tried to do with a very high density probe was to identify the neuron type based on just the electrophys phology information itself there's a preprint about this and the answer is you get about at 2/3 of the time correct and about a third of the time not if we could think of a way to do that without having to do Optical experiments with genetic labeling I mean these are really powerful tools but you know genetically labeling gives you maybe two channels of color and if you think okay there are 100 you know people just say this there are 100 neuron types in the cortex so are you going to are you going to do you know a hundred different Optical labeling experiments in order to to understand which of those hundred neuron types are part of your experiment I mean that that would be a career that one might choose it's not the one that I would choose but um it's an incredibly complicated biological space and so the struggle to me in Neuroscience is always which question are we able to answer now that gives us the most progress uh and which ones should we leave aside for now because it's just too slowo because we're still you know pharaohs with piles of rocks you know I I hate to denigrate some really really smart people but I I think life is not as exciting as it could be if there's not some other evolution of our thinking or other you know potential that's down the line that we haven't gotten to yet you know we might in in ancient Egypt not have thought about going how to get to Mars but but like hopefully there's an equivalent today that you know a thousand years from now people will look back on and and feel similarly about and even in Neuroscience like you know hopefully we've we're going to run out of stuff to do if we if we've really not just scratched the surface at this point right uh you know even if I was young I wouldn't be worried about running out of stuff to do a NEOS sign like it is really hard it's really complicated and so you know I I used to tell people look we know less than 1% of what we need to know about about mamalian neural system um and you know my friend Eve martyr at at brandise University will tell me she studies this funny neural system called the somat gastric ganglion in in in lobsters lobster stomachs have muscles and 27 neurons that that that contract the muscles and d and grind up the food and so Eve in her system knows where every synapse in those 27 neurons is absolute accurate connectivity and it will change State without her understanding why and that's because there are molecules that are changing the gain between those neurons that will and so she needs you know chemical information about this 27 neuron system and you know we're thinking we're we're worrying about mice with 100 million neurons and so it is really complicated for sure well we're just about out of time um Dr Harris thanks so much for joining us on on this episode of the podcast it was a really great um overview of of of the Gen of this really transformative technology um so we really appreciate you sharing your time with us today my pleasure to be here awesome thank you um today we have an important announcement you've just listened to the 99th episode of our podcast uh which means we are approaching a huge milestone for neurotransmissions we will be airing our 100th episode next month we're very excited to bring you another set of uh interviews with prominent neuroscientists talking about the state-of-the-art of our field um I'm here to announce today that this will be my last episode of neurotransmissions that being the 100th episode will be my last episode um so while this will be the end of the road for me and the podcast will be going on Hiatus um there will always be this sort of historical document of all the other 100 episodes or so of our podcast that we've aired over time uh that people can go back to and think of sort of as a time capsu the last 10 years of Neuroscience there have been so many fundamental changes in how we approach our work amazing technologies that have come out and uh our our podcast will be posted indefinitely so that everybody can go back and listen to these episodes if they want I can't believe we're at 100 episodes honestly I don't know what powered it through except for the fact that we've had you our dedicated listeners engaging with us over the last uh several years um thank you to all the researchers who have come through and shared their stories with us and their research careers you know think of the countless grad students and post that you've inspired with your work to carry their work forward into the future you know that ultimately was the the the mission statement of our podcast share great Neuroscience stories and career advice uh for early stage investigators and I think we've done that um very well again thank you to everyone who is ever a guest on this podcast and thank you especially to you the listener for tuning in uh it's been an incredible privilege to uh guide this podcast from start uh till now and um I'm looking forward to what MP f i has for this podcast in the future so uh stay tuned we have one more episode coming at you and I hope you enjoy it take care [Music]
2024-11-22