So it's a real pleasure to welcome today our speaker and he read Sarkar. And we were just talking before we started anywhere. It started in 2020 here at Georgia Tech. So really nobody knows who he is. So this is kind of His coming out party. I don't got his bachelor's and master's degrees in electrical engineering at the Indian Institute of Technology, Bombay. Before coming to the United States to get his PhD in electrical engineering and computer science with a minor in biology at MIT. And then he stayed in the Boston area as a research fellow at the Reagan Institute and had associated positions at Massachusetts General Hospital, MIT, and Harvard University. Before, as I said, coming to Georgia Tech just two years ago, where he's currently an assistant professor and they Culture Department of Biomedical Engineering here at Georgia Tech and Emory. And with that, I will turn it over to you. Oh, yeah. Yeah. I think it's a small group so we can keep it somewhat interactive. My title is a mouthful, but we'll get into parts of it as we, as we go on. You know, usually I like to talk, start this talk by showing this. So how many here? Recognize what this is? Okay, anymore. I mean, yeah, I think maybe this is bifurcating by age a little bit here. Yeah, so this, this is for the Star Trek fans out there. This is supposed to be the medical tricorder, right? Which in the year 2369, I believe, you know your your doctor is supposed to be able to point it at you and diagnose what's wrong with you already write just in this handheld, handheld little device. But that is not the, you know, our reality. Unfortunately, it hasn't as of 2020 to of course, medical diagnosis as of now, takes a whole lot of people supply chains and then a whole lot of equipment, which as you'll see, a lot of it is from 50 to 100 years old technology is still being used. But all of that eventually adds to delays in diagnosis. And the complexity of the instrumentation and the expertise required to run it results in high health care costs eventually as well. Now of course, you know, in the context of COVID-19, we've all become unfortunately more familiar then we want with the diagnostics issues, especially the scaling of the diagnostic pipeline, those kinds of issues. So this is the typical RT-PCR test. It starts with the swab, but then eventually has to end up in a PCR machine somewhere. And scaling this, as you might recall from now, almost exactly two years back, right? That this was, had been initially the challenge in a big way. And in fact, it can be argued that this contributed to how uncontrolled the pandemic became where there were no accurate diagnostics available in most parts of the world, right? And even here, it has been argued that if you recall, the initial PCR kit released by the CDC had issues right? Primer dimers or whatever with the PCR. And there were errors there and so on. So all of that contributed to actually what the pandemic began, became in our lab. The question, the sort of ask or the hypothesis, the broader one we, we ask is can miniaturization, right? So why miniaturization? I mean, this, this, which a lot of you might be familiar with, this idea of making chips smaller and smaller and more capable at the same time as they're becoming cheaper and cheaper as well, right? So this idea of miniaturization, can it actually do in healthcare what it has done in computation and communication, right? So can it help deliver a scalable and affordable, in this case, diagnostics to the 10 billion people world that we are heading towards. So why do we think so? Well? There's a couple of reasons. Of course, the scalable manufacturing is a big aspect of it. But also there was this interesting match of length scales now happening in electronics and biology, right? So these chips that we can now make fairly easily and inexpensively actually are at the same length scales of, of the unit features as they cells and bacteria and even viruses and individual molecules, right? So that, that gives us some unique abilities to use this chips to sense and manipulate individual biological objects. But what's more at these length scales. So the micro and nano length scales that I referred to in my title of the talk at those length scales, there are other physical phenomena happening in terms of fluids and electric fields and other energy feels. When you confine them to these length scales, certain unique phenomena happen that give you further capabilities to sense and manipulate biological objects as well. So this is broadly what we do in the lab. Micro and nanoscale tools all across Discovery, diagnosis and therapy. So all aspects of healthcare actually, and I'll, I'll invite you to check out our website to see more details there. But for today's stock, I'm going to focus on two specific stories. One is biomarker discovery, which, which goes into, you know, what to measure. So a bio marker, marker of disease. What do you want to measure? Do you know what, what should be measured, right? In that context, I'll talk about microscale immune monitoring and I'll describe what that means later. But the advantage is it brings, is you end up using very low sample volumes. So diagnosis from a single drop of sample and so on. But also you're able to do large numbers of measurements at the same time. So high throughput and large numbers of samples as well. So screening large numbers of patients, for example. All that feeds into discovering biomarkers. And then the other aspect I'll talk about is point of care diagnostics, which basically is taking these biomarkers, translating them to the clinic, being able to measure it right where the patient is in an inexpensive manner, right? And in that, I'll talk about this new transduction principle that, that we have developed in the lab using nanomaterials, integration of biomaterials, nanomaterials, which feeds into the electronic detection aspect of it, which then leads to being able to use these microfabrication techniques that I mentioned from the IC or microchip context, which leads to low cost portability and ease of use in diagnostics. So let's, let's jump right into it. So I'll start by sort of reorienting ourselves or in bringing all of us to the same page. I mean, I know a lot of this is now familiar to everybody, but just to, just to reiterate it here. So COVID-19, one of the major challenges in, from the diagnosis perspective is that a lot of people get this viral infection and don't have even any symptoms. They can still be transmitting it, but they don't have any symptoms. Then there are people who have mild symptoms. And then there is a certain set who go on to have much severe symptoms as well. So this sort of a heterogeneity of their disease is a huge challenge to diagnosis. It's not like you can say, Hey, you're infected. So for example, in HIV, let's say right, another virus, someone's infected, there's a very, very high chance that they will develop aids unless treated. So that if they are infected, you only need to know that they're infected. And you rest of your therapy or whatever you want to do. Health care is based on that. That is not so at all with COVID-19 and that's a huge, huge issue. So what you want to do is ideally that it early enough, maybe even when they're asymptomatic, you want an early diagnosis. But at that level, a huge fraction of the population is being infected. So we want it to be really scalable and cheap if you, if you want to have an early diagnosis. But then the other more important even aspect here is, you know, can you have some way of predicting beyond the known sort of correlations that we are all aware of, age co-morbidities, and so on. Can you have a more predictive or a prognostic marker that will help you to then, you know how you are apportioning or healthcare resources, who is getting more help and so on, all those kinds of things and then eventually helping individuals as well. So again, just, just describing the various infectious disease diagnostics modalities here. And I excuse me if this gets boring, like how many people here have never taken an immunology class, for example. Right? So there's a significant number of us. I belong to that category as well. So i'll, I'll just assume and describe the basics with that assumption here. So in most infectious disease diagnostics, you are either looking for the bug itself, right? The thing that's infecting the pathogen, or you're looking for the host response, which usually is the antibodies, but sometimes T-cells as well, right? So what infects you? You can try to detect it, or you can detect, try to detect the human body's response. One, as you might imagine, is a lot amplified, so easier to detect, right? The immune system generates a lot of response. The other baby a little rarer, but can still be detected depending on how large the infection burden is, for example. So in the context of COVID-19, of course, we know that RT-PCR, which basically detect viral RNA. So you're detecting the virus is the reference standard, right? On the other hand, antibody detection, it's really not useful up until things like surveys. So they tell you not whether you are currently infected or not. They are at the population scale, epidemiologic scale, you're trying to find, okay, how much fraction of my population has been infected already, right? So it doesn't have much diagnostic use currently, at least. That's because of this, right? Antibodies take awhile to come on board. This is the IgG, the most common antibody used in 0 surveys that in fact can take in month 1, but 14 days, 20 days, that sort of thing. Igm response can come up a little bit early. This is a different kind of antibody. But again, that's variable in people as well. So it's not really used for diagnostics much. It so happens that if you look at the engineering of this, right? So how, how do these diagnostics, what instrument is needed? How to scale these diagnostics? The one that works for diagnosis currently is a high complexity essay. So this is where you'll see the pooling assays and things that, for example, Georgia Tech is doing. That's to put that central lab resource, which is expensive in a central lab, but then pool a lot of samples and somehow be able to then you divide by the number of people and so on and you're bringing that falls down, et cetera. So scaling to the individual diagnosis because of this complexity is, is it stuff for PCR versus antibody responses relatively easier. They can still be challenging depending on what exactly you're trying to measure, quantitative or not, and so on. But then moderate to low complexity, I would say, in terms of the predictive nature, right? We said the heterogeneity of their disease is an issue. You want to be able to predict. None of these work basically, the viral load has been shown to be not predictive of disease course people can have the same amount of virus go totally different directions. The antibodies, of course, are not protected as of now of anything at all. So that the scientific questions which drives most of the work I'll describe today is, can we have an assay that can have a simplicity of serology? So serology says that Tom for serum based measurements, basically antibody measurements, can be having a diagnostic which has the simplicity of serology, and yet do early diagnostics, which is not possible right now, right, because antibodies come on board much later, only. Predictive theological monetary, currently, you either have antibodies or you don't. And there is no disease course prediction possible from just having it or not basically usually. But can there be more precise markers which are serological antibody based, but can be predictive as well. And yet, and of course finally, the goal is to be able to, for such a disease, the goal is to be able to scale it as much as possible worldwide. So we'd like it to be good, sensitive and specific, but also simple, so that it can be done by untrained personnel. Personnel cost is a huge portion of these costs, one is talking about, but overall low technology cost as well. So then just digging further into it, this is just showing the course of disease I described to you. And it is known of course that the asymptomatic to severe transition is correlated with age and comorbidities. It's not a perfect correlation after hospital admission though. So that's the definition of severe I'm using here. It's pretty much not known beyond that stage who ends up surviving and who doesn't is very difficult to predict at that stage. So, yeah, that's just saying that this unpredictable disease progression actually ends up contributing to patient mortality as well. What is known is it's not the virus that's causing the bifurcation at that stage. You know, you've you've gotten enough virus burden that you've landed in the hospital. It's usually the immune system, the immune dysfunction that leads to that place much more important role in the disease progression at that stage basically. So what we thought is monitoring the immune response itself may provide a predictive biomarker. And then of course, you know, linking back to the science, if we understand how this bifurcation is occurring, maybe there is something mechanistic we can come up with. Maybe better vaccines, who knows, or therapeutics and so on. So there's the science aspect to this as well. So of course, we're not the first think of doing immune monitoring and COVID-19. So all arms of the immune system that you can imagine or know of, right? Cellular, so T cells, B cells, or humoral cytokines, antibodies. All of these various people have been trying to look at these things and some correlations known for example, lymphocytes go down with severity, T-cell activation goes up, and so on. We focus on antibodies. And antibodies. There's been this sort of question in the field. Are they good or bad? Now you may say, well, antibodies are good. What are you talking about? I mean, I get these vaccines and I'm told that I'm supposed to develop these antibodies so they must be good. Well, it's not so obvious, so they are a correlate of protection from vaccines. That's for sure. It's also known that people who are immunocompromised, for example, they either lack seroconversion, meaning development of antibodies, or they have a delayed 0 conversion. So that is linked to severity. So all that is pointing towards 0. Of course, you know, antibodies are good still. But the other interesting thing is the concentration of antibodies in people goes up at severity. It is well-known that more severe disease you double up the more antibodies you have. That's sort of troubling is that they are, are they always go out there types of antibodies that are not good. So what is going on there? Why, or is it just a causation correlation thing? That you are fighting their disease more and you're developing more antibodies, but there's still good, but that's not clear. A priority for sure. And then there are some other diseases, although in COVID-19 now that hasn't been too direct evidence of this yet. There is this concept of antibody dependent enhancement. It happens in things like dengue fever, where antibodies actually bind to the virus and help it. In fact, some other cells. Because antibodies interact with cells as well. So that, that's, that's a huge concern. That could that be happening in people who sort of after having severe COVID-19 are not able to recolor. So antibodies just, again, a little bit of the immunology here. Antibodies of course, direct a large number of functions so they don't, they don't just bind the virus right day date. In fact, direct a lot of the functions in the immune system as well. A lot of the other cellular immunity is triggered by the antibodies that are bound to whatever pathogen one is talking about. So what we started to think about is that, well, it, everybody has antibodies, but our antibodies functionally meaning in these functions that they direct. Are they different and survivors versus non survivors? And maybe that is the difference there. It ends up being some antibodies are good and some antibodies are back, right? So antibodies of course have thousands of different features. And usually though, in diagnosis, when you get an antibody test, they just measure the concentration. In fact, the simplest one is yes, no. And then there is concentration, right? How much you have, basically that's called tighter. So, you know, all these other features that antibodies have pretty much are neglected and diagnostics. So drilling that into it a little bit, this is what an antibody looks like. This typical Y-shaped molecule at the front end is called the FAB and it binds to the antigen, the virus. The backend is the SCN, that's what interacts with the immune cells on the backend, sort of act as a beaker, right? Bind to the virus. Let the other immune cells know, hey, here's the virus, come eat me up or come destroy this or whatever that particular cell does, right? So what is known as this backend, although it is called F, C, the C stands for constant, it's not really constant. C.com, calling it a constant is a misnomer. The protein sequence in this part gets tune, which decides what kind of antibody it is. And then there is the sugar that gets attached, which is highly tunable as well. So subclass and glycosylation sort of end up generating a kind of a functional barcode to that antibody. So the frontend is binding to the pathogen. The backend has a barcode which tells the immune system what exactly to do with this. Neglect this, this is nothing or definitely come immediately and destroy this, this is really bad, right? So these kinds of things are being tuned on the backend of the antibody. So how does one read then this functional barcode is where we came in with these new technology ideas. What one wants to do clearly is measured by FAB, the antigen binding, right? We want to measure the antibodies that are viral specific and the FC, so the subtype glycosylation, FCR binding and all this, and we want to do it simultaneously for all antibodies. Of course, we need to be sample sparing. We were saying we'll measure all these properties, hundreds of properties for the barcode and so on. We can't be using like a liter of blood or something to get this right. These are rare molecules. So the sample use should not scale with the number of features they want to measure is a key property of the essay we want to develop here. So that essentially leads to it has to be highly multiplexed. So each measurement should give me a large number of readout solid than one readout Like a concentration or something. So we then, this came in with this idea that we should use barcoded microspheres, which can be used to develop multiplex assays. This is a commercialized essay. It, it uses basically these beads which have a Dy bar code. So it's a ratio of two dice, one, right? Diane von ir rdi, which is impregnated into the bead. But then on the bead you can go to capture molecule. And then you pull all these beads right there, barcoded. And then the barcode corresponds to what captured molecule on the B. And then you pull all these different capture molecule coated beads together, throw it into the normal assay protocols you have. So each sample is getting, expose them to all of them. And all of the different things you have quoted are binding the different antibodies. All of the measurements are happening, the bindings are happening in parallel basically. And so this, that are various commercial vendors that have, that make this up to 500 different readouts is what is claimed as possible. And then once you do that binding, you bind a probe, a secondary probe on it, which has fluorescence. Then you use something called a flow cytometer, which is basically a laser based laser and photomultiplier tube base readout of be beat by beat, right? So you've got these two class classification lasers, which classify the ratio of their dye here to identify the barcode on the beat itself. And then you've got a readout of how much was bounds. So that's a third, third thing you're reading basically. So you get these areas of three numbers and then from that you can, you, you knew which bar coated bead had which things. So you can back back out. That particular readout, whatever that particular probe corresponding to in your case. So how does one apply this to antibody profiling, especially in the context of COVID-19. So in COVID, we know that the virus has a large number of different proteins that beam system see some of them more, some of them less. It depends on the amount, amount of that protein in the virus, but also how it's presented, right? Because some antigens are exterior, right? Most of the vaccines we're getting is to generate antibody to the spike protein, which is a surface protein. But there are other proteins inside, right? Once the, once the cells get infected, they can display these antigens and the immune system can generate or respond to those as well, and so on. So what we did is picked sort of based on some curation of the literature when we started this 10 antigens in which we have some of the surface or more immuno dominant antigens, including the spike to which most of the vaccines are targeted. But then some of the other less studied antigens, which are usually internal antigens. Because we had this capability to do highly multiplexed measurements. We were like, well why not? Let's see, what's the immune system doing with these antigens. And then the other thing that is known, and I'll point out here, is there another coronaviruses that circulate in the human population. They cause common cold like symptoms. They've existed for a while, sum up to a 100 year, some 50 years and so on. So those, of course, the human population has exposure to. So we were curious. Does the existing response to this, to those other viruses? They are similar of course, and protein sequences and so on a little bit, do they matter at all that response? So we threw in some of those antigens as well. And then on the back-end. So that's for the FAB antigen, it binds to the Fc and B. We had this whole array of the antibody properties. I talked about subtypes, glycosylation, FCR and so on. So that's 17 or probes. So you can see for each sample, we're generating 170 readouts, rights for each patient sample 170 numbers eventually. And so this is just laying that out. You have the B had ICU cohorts. So basically very severe COVID-19. On day one, a blood draw is done right before it's known what's happening in terms of their clinical outcome. And then with that, with these barcoded beads, we take a drop of serum. So it's very, you know, sample efficient. Really a little drop of sample needed. And we bind it with that. We have the probes and I describe the flow cytometry based assay to you. Eventually you generate high dimensional data, right? These 170 readouts of antigens, probes and the patients basically. With that, as I'll show you, to try to understand these large datasets, you have to use machine learning type techniques. So that's what we do. And I'll describe to you what the labels here mean and how this helps us discover some active or some biomarkers here. Right? So I'll start by talking about the canonical. By canonical, I mean, this is the ones that people study more. This is the one that are in existing antibody diagnostics for example. And just to orient you to this, this radar plot which I use a lot. So each of the sort of sectors is one probe, one readout, right? And then the colors, the red is known survivor, blue is survivor. So you're seeing, for example, the non survivors have high IgG versus the survivors have IgA. So he's starting to see some differences here. A particular kind of antibody is rising and survivors versus non survivors, for example. And then you can do this with other antigen. So this was the nucleocapsid antigen, the one if you get an antibody test today, this is what will be in that test, the nucleocapsid language and it's the largest amount of antigen. The antigen that SARS-COV-2 expresses in the largest amount. So it's used an antibody tests, commercial antibody does. But some of the other nitrogen spike, spike RBD, these are the vaccine antigens for example. And I won't go into describing each of these in detail because eventually the conclusion here was that yes, there are differences. This is a heatmap of fold differences. But none of these individually us even statistically significant, right? So it's very difficult to draw conclusion here from individual univariate redox. So this is where the machine learning comes in. Again. In the interest of time, I won't go into the details of the lasso, which is a video which is down selection technique. And then peel SDA, which lets us lay out, lay this out in terms of the latent variables eventually here. But the point being made is you can start distinguishing, you can start classifying. So a minimum signature of three variables, in this case, out of the 170 that we measured can robustly classify survivors versus non survivors. So, so that's what we started to see. And this is by the way, with just those three antigen site which I call the canonical tares are still there. Coming later. We see that there is high IgA and survivors. So that's interesting. This is an antibody that goes and all your mucus membranes usually write the nasal passage lungs so that mix somewhat sense. And then we see that some of this glycosylation I talked about which tunes the backend that's changing as well in survivors. The non-canonical antigens, which pretty much no one had studied till now. We start see that there are differences in them. So you see that the red and the blue bars are different lengths in many cases. And again, in this case, again, this is that fold ratio heatmap, you start seeing some in fact, univariate measurements even are statistically significant. And sp3 and very little studied molecule. I mean, yeah, it's, it's now known what it does, but the immune response to it is hardly understood. But we start to see that the survivors and non survivors are differentially targeting this internal viral antigen in fact. And again, same kind of less appeal is DA techniques to do the classification. Interestingly, the non-canonical response, these are internal viral antigens, right, is equally good at predicting outcome. So this is sort of a surprise to us that the external antigens, it's understood that or it is hypothesized that Lee is that well, you probably are blocking the infection and that's why the vaccine is targeted to it. What are the non-canonical antigen antibody responses doing? Not so clear. But we do see some common features. They still high IgA even against the non-canonical antigens, high glycosylation as well. And then this NSP 13 thing I pointed out. So there's a high IgG 3, which is a particular subtype of IgG. We see a higher. This made us sort of wonder what might be going on here. And we started thinking, well, maybe some of these other antigens they are supposed to, they are known to be quite homologous to the corresponding antigens in the endemic human coronaviruses. Maybe that existing responses involved here in helping out survival versus non survival. But it wasn't clear yet. Then came the biggest surprise of this study. So we said, if this is an existing response, Let's put in some pre-pandemic samples behalf, right. This is before sars CoV-2 even came on board. This is from mid 2000s or something, some biobank samples that we threw them in. You know, if you've gotten anything of how these polar charts, radar charts work, you'll see that the green samples, which are the healthy, have high responses, higher than the people who are COVID infected, right? This is pre-pandemic samples. They have an existing antibody response to these antigens, especially, right, not to the canonical so much although there is some, but to the non-canonical antigens, pre-pandemic Healthy People have an antibody response. So this was a huge surprise to us. And this against started pushing us down the route of thinking, clearly the endemic coronavirus infections must be doing something here because they don't have expert, they didn't have exposure to sars CoV-2. How come they have an antibody response to these antigens, right? What is even more interesting, digging even more into it. Not only is there a response, there is a distinct response, right? So the pre-pandemic healthy, if you do a three-way classification, you can paint a pretty good classifier. You still see higher IgA in the survivors. All those properties come over from what we had there. But interestingly, we see the ****, these pre-pandemic have a higher IgM response. So the m is memory, write it just to refresh your memory here. So basically, these are memory responses. So high IgM responses against non canonicals in the pre-pandemic healthy, something we're absorbing here. So then you know this, this goes a little bit into the immunology of the mechanism here. Is it that they have these antibodies? And then upon infection, who is able to switch to the more functional IgA class survives. Those are the mechanistic questions that this ends up raising here. And then enter directly measuring dynamics. If you recall, I mentioned we throw in the endemic coronavirus antigens directly as well, right up until now I've talked about canonical and non-canonical sars, CoV-2 antigens. Now these are the antigens from the endemic coronavirus. We see that even the endemic responds to the endemic coronavirus antigens only. So if you don't measure the response to the sars CoV-2 antigens at all. You measure response against only the endemic coronavirus is, you see that even that response can classify non saliva versus survivor, right? And then this is sort of a motif emerging here. Higher IgA response, we keep seeing it here. And then we also see that the pre-pandemic healthy actually corresponding to the ER compared to the eventually, those who got sars CoV-2 infected have a higher antibody response against the endemic coronavirus as well. So that makes, makes one sort of wonder, well, maybe having a higher response already to the endemic coronaviruses helps you, in terms of survival, are not getting that infection, right, not clear yet. So that sort of brings me to the summary of this part of the talk. So we think this is a biomarker of survival clearly uncovered by the multiplex antibody profiling pipeline that I described to you. This is from day one sample, just to remind you, right? So when someone's going to the hospital, at that point, this readout can be taken and potentially helped in apportioning healthcare and so on. Canonical and non-canonical, which canonical form to which the response is well studied? Non-canonical, not so studied. Clearly they need to be studying is one of the conclusions here because the responses to them even can be predicted. But and also the endemic coronavirus responses are productive as I showed you. And then the, the biggest surprise, as I said, that the pre-pandemic ****, they actually have IgM response to the non-canonical sars CoV-2 antigens. So with this, you know, in terms of the mechanism B, B started thinking that maybe the switching of the existing IgM to the more functional IgA and IgG 3 is involved mechanistically in, in being able to survive eventually the severe infection. So we think, you know, in terms of a biomarker, definitely this is useful, but maybe this is also, has also uncovered something that these antigens might need to go into future vaccines, right? If the mechanistically correlate in terms of helping out in severe. So maybe, you know, to, to prevent, prevention, to prevent infection or development of severity et al. The current vaccines or maybe good enough. And that's what we're seeing even in the face of these, all these variants coming out. But post severity, maybe there are other antigens needed. So that's sort of showing the power of this pipeline or this workflow as well. Where now it's a discovery engine, right? So it's not only biomarker, but new science as well and maybe potential vaccine antigens and so on. So with that I'll, yeah, so that's that's our pre-print. I welcome you to look it up. So we have pre-printed already what remains and review. But in the rest of the time remaining here I'll talk a little bit. You know, since this is nanowire texts, I should go into the nano domain a little bit here. There's been a too much biology here you have for some people. So I'll talk about this point of care diagnostics and transaction technique that we have developed, which is, which goes into then can these multi-varied biomarkers, these fancy things I talked about, right, 170 readouts. What use are they if they cannot be measured inexpensively in the hospital or wherever, right? So ideally at the point of care. So eventually the current readout methods, all these lasers and photomultiplier tube, the great, very sensitive, very accurate, very expensive, right? So and then the manpower, as I said. So the trained manpower required to run them comes at a cost as well. So these things make this, you know, it's good to know for scientific research, great but difficult to translate. So our thought has been, you know, can we eventually take all of this and build it into a little chip somehow, right? That's the science fiction vision currently. Like, you know, take a drop of blood put into that chip and somehow on your phone comes a readout, right? That says, oh, you know, you need to get better care because you are at risk of developing very severe COVID-19 for example, right? So this is this idea of no portable and inexpensive point of care. So allies as other kind of essay which, which are used to measure antibody, various immuno binding. So you want to develop something like this. So for this, we started looking at a lot of the cost, will look at, you know, what adds to that in terms of the technology and the instrument cost? What ads most? Well, it is this lasers and photomultiplier tubes and expensive optics involved, for example, right? And then there's mechanics associated with them and so on. So we thought, and because I'm an electrical engineer, that's one of the bigger reason as well. We started thinking, you know, can we develop can bypass all of that, right. Can go directly to an electronic readout. I mean, you know, your phone is I electron has the electronic input. Let's generated directly and electronic read out somehow, right? So we found this particular chemistry. Usually it's a very similar chemistry to what is done, what I told you about on the beads and so on. But on the beads it's a fluorescent readout eventually, right? So there's a fluorescent tag eventually on the secondary label. Here it's an enzymatic tag which can do silver reduction, right? So you can see what I'm, where I'm going here. The silver reduction can deposit metals right on the surface. It's selectively bound to. This has been used for electron microscopy and so on for example, because silver is sort of, It'll, it'll be visible in a particular way under an electron microscope being a metal. So that is where this has been mostly use. But we started thinking that this idea, which is called enzymatic mineralization, which has existed. Can we sort of try to build a direct electronic detection technique using this, this method here. So first of all, of course, we wanted to see whether we can even couple of these to an antibody assay, right? So that, that's relatively simple. You coat whatever the chasm capture molecule you put your samples on, put the enzyme probe on, and then the enzymatic substrate, it deposits the silver that seems to work right there, just little wells on a glass slide. Healthy samples, not that much silver, COVID positive samples, a lot of silver, you see a dilution curve. So that's just the sample being diluted, right? So you see the different sort of visually, you start to see it already. But what about this electronic detection idea then? So we thought, okay, so there's the silver deposition happening. We're going to put in some electrodes there. Some electronic property will change conductivity, resistivity, whatever. And we'll have it. Well, not so fast. So we made the chips, of course, yeah, right here at IN, it's just very simple goal, microelectrode arrays, IDs, fingered electrodes. We put that chemistry on it. We see silver deposition, but it's sort of, you see how it is sort of getting repelled by the Golden almost right? So there's silver deposition happening, but it doesn't seem to cover the electrode somehow. So we were very, very curious what is going on here. Of course, you know what this ends up meaning is in terms of the electronic properties. It's not very repeatable or changing that much, right? So from there, looking into it in detail, we see that. So this is this is the chemistry I just described to you. But we start to see that yes, there is medicalization and then looking at it under the electron microscope reveals something very interesting. We see that there is the silver nanoparticles getting deposited, right, the enzyme coating, but supposedly everywhere. But you clearly see that there's more silver on the electrodes versus that glass, right? If the glasses where we want it to be in order to bridge these electrodes. So that clearly is the reason it doesn't work somehow, right? And then you can zoom in and see that the gold has a very large amount of silver. It's almost a very dense silver on the gold electrode, but on the surface here, not that high. So of course I'm not digging back into literature. We be realize that of course, of course gold also can catalyze silver reduction is, is something that is known. And that's what might be happening here. It had been said, I mean, it's been said in literature as well in a goal has to be at, in a sort of a nano material or nanoparticle format to be able to do that, gold electrodes shouldn't be doing that, but they are doing it for us. So, so then, you know, sort of confused here. Nevertheless said, Okay, let's put in some gold nanoparticles in here. What will happen if we add some gold nanoparticles instead of the enzyme, right? So there's no enzyme, only the goal, it seems the gold does much better than the enzyme. So forget the enzyme, Let's put in the gold, right? So now the tag is only gold nanoparticles. We see that we have mineralization happening, but it's still pretty low and it doesn't conduct. In fact, it's lower than the enzyme case. So clearly the enzyme works better than the goal. And the goal does something in the nano format as well. But the best thing is the gold electrode somehow, right? It's generating all this silver almost as if it's stealing the silver from, from where it needs to be, right? It takes on all of it and there is not much getting deposited here. So with this, we came up with this idea that, well, what if there are gold nanoparticles and enzymes both the right? What happens if you do that? And that was sort of the aha moment when we tried this. So what you're doing now you have the nanoparticle probe, but you also have the enzyme probe. And suddenly you see a lot more just visually, a lot more silver deposition, right? So this is not the addition of this in this even visually you can see it's a lot more so some synergistic effect is happening here, which makes it a lot more than what the enzyme could do alone or the nanoparticles could do alone. They come together to do something different. And you then start seeing in the electron micrographs, there is a dense silver carpet everywhere, right, on the electrodes. It always had B. But now on the glass as well, zooming in, you see a connected layer, a more connected layer forming right? Instead of these disconnected particles. So that told us that the nano structuring of the surface by adding this gold nanoparticles clearly generate some kind of synergy with the enzyme, which we don't fully understand yet, the mechanism off. And if one of you does, I'm definitely love to talk to you after. But this definitely seems like it would generate a signal for us, right? Because this is looking continuous layer here. We went in and measured, of course, the impedances. So this is just telling you that when you have only the enzyme, you have a high non repeatability. As I said, it gets us somewhat conducting, right? I mean, the open circuit resistance of these things is above a 100 megaohm or something at this frequency. So you see that it's coming down a little, it's non-repeatable. If you have just the gold nanoparticles as I showed, right? It's not conducting et al, There's some silver but not conducting. But when you add these two things, it becomes fully conducting. It's almost a short circuit basically, right? So from that 100 megaohm down to like less than 10 ohms. So 67 orders of magnitude of impedance change. If you do it this way. So this, we thought, hey, this is our electronic transaction technique which can help us. We've seen that it can be coupled to an immunoassay. Now we have a way to turn it using the gold nanoparticles. Now we have a way to turn it into an electronic readout. So let's run with this, right? So that's exactly what we did. One question is, is this quantitative? Is this only yes, no, right, one would want a quantitative readout of course. So what we've established here that yes, it's quantitative. So the way to read these plots, if you've never seen in lies a plot before. This is a sample dilution, right? So highest concentration, lowest concentrations of one over this is the concentration kind of samples being diluted. Right? And then you see that there is a quantitative readout happening. The blue versus red is basically just a matter of detail. Two ways of adding the nanoparticle. You can add it after you have bound sample as a probe, right? Or you can even just decorate your surface already with the gold nanoparticle before even starting the assay, we see that decorating the surface ahead of starting the SA gives up two orders of magnitude higher sensitivity, right? And this is just showing you the difference of the silver as it goes down. And these other controls which basically don't have much conductivity at all. So here, of course, you'll notice that I flip the scale here to make it look more like the Allies occurs that the biologists are used to seeing, which basically goes up instead of down. Of course for us that resistance actually goes down. But yeah, so here's the 10 ohm and a 100 megaohm is down here. So you see a nice dilution kind of happening. So that establishes that a quantitative electronic readout is actually possible. Of course, once your micro fabricating, this is just trivial, right? You can build a lot of these on a single chip. This is what we're talking about in terms of how computation becomes, became cheap and all of that good stuff. So we of course talked to do that. This is a single microscope slide, one inch by three inches. Lots of electrodes on it. I believe it's 16 different. And you can see the visual readout is sort of there as well, right? Hi silver, silver COVID positive versus healthy. We started calling it easy Eliza. It's an acronym. I can talk about it if you're interested. But the Allies are the gold standard technique what it sees in terms of the readout. So here's optical readout, right? Using the expensive instrument that I talked about. And you see that there's good concordance, definitely. But also much larger differentiation here happening, which leads to a slightly better sensitivity specificity. This area under the ROC curve here, right? This is because of the switch type characteristic of this asset, right? Because it was not conducting, it became conducting. So huge difference suddenly gets created. So that tells us that you can do high-throughput electronic readouts using this. Then the last thing to be able to do those multivariate markers that I talked about. Can this be multiplexed? Do I need to take many different samples in this case you, so they're over there, you know, there was the barcode and the bead and so on, laser and everything. What about here, right? So the question becomes, well, we know that it deposits silver, where the sample and nitrogen had bound. How local is this is the question we asked. Like if we had a small spot which had a positive sample of one antigen and the next, next, right next was another antigen. Would there be two independent redox or would they mix up, right, because the silver deposition is happening from a liquid substrate, does that reduce silver from one, go to the next? Or how does that work out was the key question we were trying to ask. So we built a very simple, reconfigurable sort of, you know, there are four electrodes here, very simple chip flies. There's a RACI level, sort of plastics or PDMS layer. Essentially we're making wells. You pattern for antigens. One of them expect it to be positive, three negative. And then you of course then want to put a single drop of sample, right? That's the idea. Can you do this measurement from a single drop of sample is what we want, right? So that's what we did. And then you see if wherever you put the positive mark, that's where the load resistances tax where the high silver race and the others remain non-conducting, right? So a 100 megaohm versus less than 100 Ohm. So that tells you that from a single drop of sample by patterning different antigens, you can get a microarray type electronic readout here, right? So that was pretty cool. We wanted to apply it immediately, do something interesting. And in this case, the easiest available sample at that constant, this was a little bit later. Now in the pandemic, right? Much more recent data. So you're like vaccines are out. So let's look at how the response to a vaccine versus people who are COVID infected. Is it different? And that's expected to be different as I've talked to you earlier, spike antigen is the only antigen in the vaccine. People who get infected end up developing and, or responds to many antigens. So you know, we, we put in the spike and then you will to capsid nitrogen here. The probe side also be multiplex here. So that's one of the beauties of this platform. You can measure IgG and IgM, right? So by putting different probes on the different, well, so that's what we did. And you see that of course, the pre-pandemic healthy, they don't have in this case any response. So these are, these are sars CoV-2 antigen solely. So of course, in all that other stuff, I talked about dynamic viruses, it's not relevant here. So they don't have a response at all. People who got infected have responds to all the antigens. And in this case it seems they still had an IgM response as well. So probably it was sampled early in their infection. And then the vaccinated people. This happens to be from my blood actually. But back then that was all we had access to. But you see that you have IgG spike antigen response, but no response with the nucleocapsid, which is not in the vaccine at all. And this tells me that I was probably not infected right? In them. Because again, in this case individ this disease, you don't know whether you are in fact, unless you get so this is my antibody testing me that I only have a vaccine response. Igm is not here because this is two months after, after infection. So the IgM response, It's probably quenched by that, right. So with all of that, you know, tying it together, you have this chips and electronic detection happening, multiplexing and so on. Of course, this is VD finally enables it being pushed into translation, right? The idea is, it's a simple electronic readout. You don't need. It's just resistance change. You don't need much to detect it, right? It can be done with a multi-meter. A lot of data I showed you is with those handheld yellow multi-meter is that all of us have used an undergrad labs and so on. But we built a little portable OCR meter of our own as well, just to be a little bit fancy, I guess. So this is, this is fairly simple electronics. There's a microcontroller in there, a single chip and CR meter, I think from analog devices in their interface as well. And you can just run your essay with a drop of liquid on it, plug it in. And then there's this app, one of the undergrads in the lab, the lab, which can show you a readout showing these two are positive, those two are negative or generate a graph or save the data to Google Drive and all those, all those good things it can do. But this basically now is getting to a point of care electronic antibody profiling. Of course there's work to be done, but I think I've shown you that some of this is starting to become possible here. And so the next thing actually we are currently working on, which is more, becoming more relevant now because the variance now lead to that much severe infections, right? But monitoring vaccine efficacy is a new need, right? Each of us would like to know, is my vaccine working right? Or do variant came out? Is it, Is it working against the BA to my vaccine? Is it's still working? We want to know but there is no such as a, I mean, yes, in the research labs there are assays, but they are so difficult to translate that most of us don't even know that these assays exist, right? So we're working on that. Developing individual electronic vaccine efficacy assays. We're looking beyond COVID, as I've shown you. Wherever this antibody functional characteristics like ans and so on can track with disease progression. You can apply the biomarker discovery and detection techniques. We have some preliminary data showing in diseases like bacterial diseases like typhoid. So this is a viral disease I talked about, but bacterial diseases like typhoid, it. These markers can work fungal as well. Also some completely non-infectious infectious contexts. But antibody driven rejection of graphs is a huge problem with people who get transplants. So we think we can be able to predict earlier whether a reduction is occurring by measuring these antibodies and like ants. And of course eventually, you know, why just antibodies, why just plugged, right? Because it's a simple immuno binding assay that does not really depend on the specifics. So we think we can be more ambitious and see, oh, can we compete with PCR, right? Can we be so sensitive? Pcr has a tuple, the 40 amplification, right? 40 cycles. So that's a large number. Can we be so sensitive to displace PCR as well? Now, that's a tall order. We are far from there, but that's something we want to try. And of course then larger objects. I think this should actually be more trivial and easy, but we haven't yet done it. Cell-based assays, bigger objects, something binds, you can detect it. The idea, so why not cell-based assays rights, l towns, amino phenotyping, functional assays, all of that. Sorry, I'm running out of my time here for sure. So I'll just acknowledge that team, some of whom are here in the lab, small but growing team. And of course, none of this can happen without the clinical collaborators who provide the samples and the understanding of immunology and so on. And of course, acknowledging my funding sources here. So with that, I'll take any questions. You might have questions from anybody here in the room. Yeah. So the question is the microfluidic spark health. So obviously there are probably mixin constraints on the time response. In maybe even there are some, I don't know because maybe some crossover between different native cross that can influence the response. Can you comment on those kind of engineering now, get in just an inspection technologies how important and decide what to pay attention. So I think yeah, awesome question and that's something we worry about a lot. We've been lucky a little bit that the crossover effect is not so much there. Mixing retained in terms of improving the repeatability of some of these assays, especially down at the very low picomolar and so on. We think that could start mattering. We would like to do it. We haven't currently done it, right? Or do we want a microfluidic mixer separately or some electromagnetic mixer or what have you. We don't know yet, so we would like to go there. Haven't gone. Great. Hi. Great talk by enjoyed it. I had a question about something you mentioned earlier. You said that the impedance measured at that particular frequency was. Satisfactory, but I realized that maybe you're running things at different frequencies. Yeah. So we found that there's some paths for impedance Piotroski p type of approach or this curious where you frame on the light. So that if I tell you too much, I'll give away some of the results. Newer ones which my students may not like. But definitely, no, I'm just kidding actually. So, so the thing is, you know, this is a resistor. Eventually I got two ends of a varchar, right? So, so we did a whole spectrum, but they'll see our meter and that's why the portable thing is an LC army there because back then we didn't know what will happen. So we started developing a portable OCR. Turned out that it's mostly a resistor, right? So we see a clear capacitive, the kind of thing you would expect a spectrum versus a bear resistive spectrum now, but, but does that mean that it can be converted into something that will be more sensitive if you do a full spectrum, that's, you know, what's the space type of thing? It's coming. Hi, excellent presentation. Thank you. One other question I have is that in your sandwich I say, what does see a fluorescent tag. And number two is, did you exert any photobleaching associated with when you use that laser? Yeah. So I think in these so maybe you are others here are aware of the Luminex platform. That's what this is. The classification lasers. I think our rhodamine or something, some IR die. And those are pretty high amounts inside the bead. So those, it doesn't matter much. The backend probe read out is just phi co-editor in that it does matter. So one has to be careful to not expose too much to ambient light and so on. But eventually, you know, within the limits of differences that we want to look at, it seemed to be fine is what I'm able to say here. Thank you for your talk at a couple of questions. The first one is, do you immobilize your antigens onto the glass part of your chip. The second one is you've, you've shown some results of sensitivity of your device for different, different dilutions of your sample, or how does it translate into actual concentration of antibodies and what is the limit of detection? And the last question I had for you is, okay, once you put a drop of sample onto your your device, I understand that there's different steps of washing incubator and with other arms, samples and stuff. How long does it take actually the entire process. Thank you. Yeah. And those are all great questions. And there's a lot of what I would say remaining to be done in some of them that immobilization question, let me get it out of the way. We use this molecular attention layer standard one called poly-A lysine is a positively charged, so it's a charge-based thing so that, that just works for us, at least items with the sensitivities. So now with clinical samples, it's a little bit difficult to define how much exactly antibody they had in terms of nanograms from. It can be done by running a calibration essay, right? For example, I can separate allies, advocates stand out or something. But the clinical samples, I don't know. I do know that our sensitivity car again, this is based on data I'm not shown here, is somewhere in the middle of, you know, what Eliza instruments can do versus what your lateral flow assays, which just do a yes, no. So we have some pitch somewhere in the middle there. We'd like to go towards the full Eliza, Right? But we're not there yet I think. And eventually the point, I think becomes we get lucky here because antibodies are such enlarge amount in the clinical serum that we're able to work this out. So the bead-based assay I talked about, right, for the immune immunology and the biomarker detection aspect, that's our services and that's not even enzyme amplified and it still works right now. Of course that's with the laser. So what that tells you is that this is, for this application seems good enough By we need to find out what the exact limits are and improve it. And then there was a third thing was oh yeah, shop, shop, right? So that's where the microfluidics comes in and we'd like to automate it. Currently. It's long. Like let me just be frank. Because the incubation with the sample itself can be half an hour or an hour and then there's washes and so on, basically, yeah. Okay. I think we are at the top of the hour, so I'm gonna take road one more time and thank you for coming. Thanks. Thanks to you. And I'm here. If you have other questions.