All right, well, I am going to talk a bit about some precision medicine tools. It's gonna seem like a smorgasbord of topics. It builds upon one of the topics for this session and then some of the overarching themes that you'll hear at SSCM this year. And so I hope you enjoy it. So I have no significant financial disclosures to report. We do have a patent pending for whole blood Eli spot. I will talk about Eli spot today, but Eli spot is a commercially available test. We don't have any rights to the test itself. We just have a patent pending for an aspect of whole blood Eli spot that we've been utilizing. And I'm a member of Immune Functional Diagnostics. I don't receive any direct financial compensation. Certainly could receive in the future financial compensation. So we've already discussed sepsis as a threat and the outcome and morbidity associated with this disease and syndrome. We've talked about the nonspecificity of this syndrome and perhaps trying to find ways to try to become more specific. We've discussed a little bit about the muddled divergent host immune responses that become maladaptive and muddled, both in the innate and adaptive immune systems. We've talked a little bit about what happens in sepsis with the early responses associated with innate and as measured previously by different cytokines, including TNF alpha and IL-6 and activation of complement and coagulation of phagocytes and some of the cellular immune responses. As well as we've also discussed briefly, Jamie mentioned about the decrease in HLADR expression and how this has been partnered with LPS-stimulated whole blood or TNF alpha production decreases to demonstrate paralysis of the immune system and how these patients over time that have suppression of their, especially their adaptive immune responses are at high risk for secondary infections and decreased recovery or paralysis leading towards further deterioration. But it's interesting because you think that if you leave the hospital, that that's a great thing that you're going to do great and you're going to do wonderfully. And so this Kelly Clarkson song, if what doesn't kill you makes you stronger, may not be true. We don't, at least if you look at some of the data associated with sepsis, because we know that there's a post-sepsis syndrome. There's especially in trauma and trauma-induced sepsis and other variants of sepsis. Patients that leave the ICU have changes in their protein metabolism and their immune dysfunction. They have fluid retention, cognitive changes, recurrent infections, and have a higher likelihood of death within the first six months from sepsis than any other condition. And this persists for not only months, but for years likely. Down at the University of Florida, Phil Efron's group, were able to actually characterize persistent inflammation, immune suppression, and catabolism syndrome, PICS, in these sepsis survivors in these three clinical trajectories. Those that have persistent inflammation, immune suppression, and changes in catabolism. Sachin Yendi's group at Pittsburgh did this really nice. And I think, you know, it seemed at first some simplistic because he looked at specifically CRP and soluble PD-L1 over time after admission and demonstrated two common phenotypes in this study that came out in 2019. He demonstrated that about a quarter at three months, at six months, about almost a third, and at 12 months, still about a quarter, still had elevated nonspecific high-sensitive CRPs. And additionally, many of those individuals had elevated soluble PD-L1s. And so by identifying both the hyper and hypo-inflammed group, this group was demonstrated to have increased worsened outcomes and increased risk for death. But it still didn't answer the question that most of us at the bedside really want to know when we discharge someone. Who dies? Who lives? And as I mentioned earlier, many of our patients will go off to different skilled nursing homes and different places than when they came to the hospital. And who lives to die, waiting to come back to the hospital with a secondary infection, especially when an insult like sepsis ensues. And then furthermore, how can we translate this down to the individual patient that's sitting at the bedside? Precision medicine. We like this buzzword. Well, this Ice Cube movie, are we there yet? Where are we? Where are we currently going? And how do we get there? So I think it's important to understand the definitions because we use these commonly interchangeably, but probably erroneously in many situations. So I just thought it would be a good idea to at least just briefly go through these so that we're all talking the same talk. So precision medicine is when diagnostics or treatments are optimized for genetics, lifestyle, and environment. But it's personalized medicine where treatments are prioritized to a patient's needs and preferences. Endotypes are biological subtypes defined by distinct pathophysiologic mechanisms within a phenotype, being clinically observed characteristics that characterize a group of patients within a disease or syndrome like sepsis. And then we will talk a bit about biomarkers that are agnostics and omics technologies. So let's take a look at history. And Maslow had a great nature medicine paper, I thought in 2022, that highlighted some of the history of precision medicine. And certainly when we think about critical care, a lot of the foundations of what we currently see in the bedside were described from the 1950s in the post-polio period through the first description of ARDS with the development of biometrics or continuous vital sign monitoring, understanding of support for specific organ systems, being able to have advanced hemodynamic monitoring, and then utilizing these clinical biomarkers, if you will, to be able to develop objective scoring mechanisms that could be used from clinician to clinician and in research studies to be able to define specific metrics for success. And after these foundations were initiated, we accelerated the use of these with first, furthermore, expanding upon our definitions of ARDS, of developing a commonly used in 1992 definition of sepsis, and then building upon that second definition of sepsis all the way through SEP3, and certainly in ARDS and Berlin definitions. And we've accelerated our ability to take in different points and discriminate from different data points to develop and accelerate our ability to get to a point where we can precisely evaluate patients, but we're just starting. Precision medicine, in my view, at least in sepsis, it's really only just began in our field, and so we've got a lot of hurdles to at least overcome before we can truly get to personalized medicine for individual patients. What's the hope? Well, certainly we want that holy grail test that will make that diagnosis, that will identify severity of disease, that'll help us understand who's at risk, who has that specific polymorphism that was just mentioned that could help us understand who's got a high severity of illness, perhaps for a specific entity of sepsis, and then understand who's at likelihood for further deterioration with good accuracy, but rapid turnaround. We want that test that's gonna come back that we can actually intervene upon in a meaningful timeframe, and yet certainly this is not where we're at today. We certainly have come further along than we've been, but this is certainly what we hope for. And so I started our conversation today to talk about, we're talking about sepsis, and yet a lot of times at the bedside, this nonspecific clinical syndrome, we still worry about whether or not we can say someone truly has sepsis or not at times. It's unclear with that patient that has persistent fever with an unknown source or unknown etiology, and so sometimes at the bedside, we're left to decide, does this patient really have sepsis? They're not in shock. They certainly have the nonspecific SIRS criteria, and I wanna understand if they truly do have sepsis, what's their risk for further deterioration and mortality? How can I actually put them into a low or mid or high sepsis mortality risk? And so previously, we haven't had great tests to be able to make this diagnosis for us. This has been a clinical diagnosis, and over the last couple of years, there's been a number of different agents that have come across the market for precision diagnostics for sepsis, and in my view, they're still in their infancy, but they're one step closer for us to get to hopefully where we're going to go, and those include some new technologies from Inflammatics and Immune Express who specifically use transcriptomic approaches with machine learning algorithms to score the likelihood of infection, whether that's bacteria, virus, or disease. Why is this important? Well, previously, we've had difficulties and hindrances for identification of sepsis because many of the tests that have come before these transcriptomic approaches have had low sensitivity and specificity of diagnostic options. We've used and relied on nonspecific biomarkers. In fact, we've been told we need to use these biomarkers to measure patients with sepsis, including lactates or procalcitonins or C-reactive proteins, and we've used more nonspecific objective measures with clinical variables like QSOFA scores and SIRS criteria so that we can hopefully move towards a diagnosis of sepsis. Our cultures have been not positive frequently. Less than one-third of all patients that have sepsis in the United States and outside the United States have a positive blood culture. They're prone to false negatives and false positives, and it takes up to 48 hours for turnaround time. And so these have been some hindrances in why we need to get towards rapid diagnosis first and foremost, even before we can talk about the biology, I believe, of sepsis. And I'm not going any further here. There we go. So Inflammatics uses a proprietary 29-messenger RNA transcriptomic metric, and they've done a number of tests with the using machine learning to be able to identify whether or not there's the presence or absence of infection. And then in this study, you can see here that was published by Scott Brackenridge now at University of Washington in JAMA Open in 2022, that in the single-center ICU of 200 SICU patients with suspected sepsis, that this specific transcriptomic was able to estimate with some good accuracy sepsis in 30-day mortality. And furthermore, we're able to demonstrate that their proprietary IMXBVN3 bacterial metrics was able to see some small serial increases in those patients who subsequently would develop sepsis. And so in this study, this at least lent them some support that they were on the right track to develop something that could be relatively used soon in clinical practice. They further went on in that paper to look at the prediction of secondary infections and adverse clinical outcomes. And you can see here in the single-center ICU after they collected their whole blood within 24 hours after the clinical diagnosis of sepsis based off CEP3, that they were able to have at least a pretty significant demonstration of accuracy in prediction of those secondary infections. And then furthermore, they took their inflammatics test and wanted to use it in context of emergency room to back it up into where we really are trying to understand if someone is truly septic or not septic. And so they evaluated this in almost 400 future surgical ICU patients with suspected sepsis or infection. And they looked at 28-day mortality and their test was able to demonstrate that with a severity two classifier, they could predict 28-day mortality and seven-day intensive care unit care with pretty good accuracy across their test. So this test for inflammatics previously was in development. And then back in the last November, they were able to get an in-development FDA clearance to be able to use this in clinical care. Septicite and Roy Davis is in the audience here and Septicite is also sponsoring the select lounge here. So we're most appreciative, has a really exciting technology. And this is just one piece of their different devices, but they measure with their mRNA and white blood cells and quantify in their two white blood cell genes. They're able to come up with a sepsis score to be able to see the likelihood of sepsis and be able to differentiate it from non-infectious causes with pretty good reliability. And so their score goes from zero to 15, which would demonstrate the likelihood of sepsis. They're able to get results, and this is previously they've got even faster than in about 60 minutes by looking at their gene expression assay, measuring the ratio of mRNA of two host response genes and get it in, I think now he could tell me if it's, is it five minutes or 30 minutes, Roy will tell us later which one it is with the new FDA rapid clearance Septicite test. But this is just one more way that people are developing at least precision medicine tools to be able to demonstrate the presence or absence of sepsis. And here's the AUC for their Septicite rapid in differentiating sepsis versus SERS, it's 0.85. And when you compare this to lactate procalcitonin and then septic score, you can see that there's a much higher AUC with use of Septicite rapid as compared to these nonspecific markers. What's great about Septicite is this is FDA cleared and can be moved into clinical practice. And so there needs, in my view, needs to be some further clinical validation for different contexts, certainly patients with oncology processes or children that might have not just sepsis but may have things like malaria or other perhaps inflammatory diseases like HLH. But certainly we're on the right track at least for initiating the ability to make the diagnosis and differentiating yes, has sepsis or does not have sepsis by use of these two transcriptomic approaches. So precision in knowing the bug, you know, first of all, Dr. Wurtenberg gave a great talk specifically about precision in understanding the pathogen and host interactions. What about whether we know the bug or not in sepsis at least clinically, does it matter? Well, we've commonly thought forever that it didn't. I'm not sure that that's truly true. Even though there's a final common pathway, the host immune responses are certainly different by different organisms. When we looked at the literature over multiple studies as far back in 1995, all the way through the young 2000s, when you compared culture positive versus culture negative sepsis, the outcomes were not very different between both groups. So one would surmise that it doesn't matter the name of the organism. But other studies have disagreed with that and have actually have attributable mortality. And certainly as you move towards KPC and ESBLs, certainly these organisms confer a much higher mortality. And so in my view, it's important to probably know what the organism that perhaps may be potentiating your underlying host immune response. And so a number of clinical centers have certainly moved towards more precision approaches. Our site and many other clinical sites have moved towards matrix assisted laser desorption ionization time of flight mass spec, which can frankly take away a lot of the human error that occurs with contamination in a controlled environment, at least at our institution, our robot controls all of these samples. No one has to look at these in the middle of the night. They turn positive by using this and you find results in a much faster fashion than previously done. And so MALDI-TOF is actually, I think, one step at least in getting a name for some of the bacteria that perhaps may be inducing sepsis that you're seeing clinically. So moving from diagnosis to disease evolution, we've talked a bit about a battle that's gone on for years to our preclinical studies model humans. And I do agree. I think mice have certainly gotten a bad rap. I do think that there are some nuances specifically in trying to translate certain mice experiments because I do mice, let's say, that are, for instance, my plaque six mice that are six to eight weeks of age, they're all pretty much the exact same model. Obviously I can modulate their genomics in a fashion, but they don't represent perhaps what we see out in the population. Now, that doesn't mean that we can't, we're not informed by some of the things that we're able to decipher from different mechanisms, but we need to understand is what are we getting from the mice? Do they really need to translate humans? And I think Dr. Wardenburg's points are well taken that in fact, there's a lot of information mechanistically that can inform future human studies. And in fact, they don't actually have to recapitulate the human experience if they can inform us about certain pathways that may be relevant then to look into these heterogeneous populations. The other thing that I was most struck about from Dr. Sturgill's comments were if you look at patients in Kentucky that may look exactly the same as people that look in, for instance, where I live outside of Cleveland, what are the underlying genomics and how do we actually recapitulate in our clinical studies genomic differences between the populations, but also environmental cues that may influence some of those expressional differences? And that's challenging and I certainly don't have an answer for that, but I think certainly something important that we need to think about when we think about our mouse to human translations. So how do you find precision in a syndrome? Why did this patient get this disease at this time? Why did this patient have a different outcome compared to another similar patient? Well, there's so many problems with adopting, answering that question to sepsis. We've talked a lot about that heterogeneity that principally barriers our ability to find efficacious targeted therapies. We've talked a bit about heterogeneous populations. And so Chris Seymour at Pittsburgh had this really nice, I think, review in Critical Care of 2017 that looked at three potential scenarios that may be challenging to our ability to develop precision medicine therapies. And he said that there's one that includes no multimorbidity. And so doing a study in this group is nice because each cluster has the same set of comorbidities. Doesn't explain any environmental cues or other perhaps underlying genomic expression differences, but these at least have the same set of comorbidities at least at time of inclusion into a study. But then when you start to look at comorbid conditions within a specific group, you can see that these comorbid conditions can help form the discrete group, but others may actually lead towards the discrete group, further complicating the heterogeneity. And furthermore, we don't look at those patients that leave the hospital and follow them serially in many of our research studies. And so understanding these patients over time is gonna be relevant to our ability to develop new precision medicines as people are actually leaving the hospital. We have a challenge with how to generate real-time data. So this is a publication that should be coming out soon where we looked at all of the different potential causal mapping issues related to understanding how to come up with precision medicine in sepsis. And as one might imagine, there's a number of feedback loops that certainly are relevant, but it's a complicated map that's not just as simple as we're going to modulate this, which will affect areas down here. Our other inherent challenges are this, data, data, and more data. We work in an intensive care unit where we've got continuous variables, sometimes not even being captured amongst patients who have different clinical laboratories and imaging studies, and perhaps can have research laboratories that are derived. There's a volume, there's a quantity and a dimensionality of this data, and the speed at which this data comes in, and then the patient certainly changes, and then the condition changes. And so evaluation of this data can be complex. And then how do you harmonize across patients similar data sets? What's the quality and reliability of the data, especially with the heterogeneity that exists across most of our ICUs when comparing our ICU to your ICU to someone else's ICU? And then for critical care, is the data able to be analyzed fast to alter the syndrome or disease trajectory, or does it even matter that we have to analyze it as quick as we think we need to? Because if many patients are specifically able to be dichotomized into suppressed versus hyper-inflamed as Sachin had demonstrated previously, perhaps what we need to do is understand those at the highest risk for decompensation relevant to at least their, what appears to be not mechanistic, but more descriptive phenotype generation. And then we've got different types of data, static data, biometrics, as mentioned. We've got the ability to wear a watch like I have right here and most of you do as well. When you leave the hospital, you can certainly look at things like your heart rate variability. You could potentially look at some ECG findings because these are FDA approved for ECG reading for atrial fibrillation. And so the ability to be able to take a patient that leaves the ICU and can wear a watch and wear a t-shirt that actually can tell you the temperature and can give you some clinical cues that demonstrate that they're having decompensation certainly has not been realized yet with at least following up patients at high risk for sepsis decompensation once they leave the hospital. Let alone omics. We are in a world of genomics and proteomics and metabolomics and lipidomics and epigenomics and transcriptomics. And all of this data is certainly relevant to understanding the underlying biology of the syndrome. But how do you actually take it in context for the patient that evolves specifically early in their disease course and then over time and then has environmental cues that may alter some of these omic differences. And then a number of different tests on the market that you can use in the research laboratory that frankly are not available in the clinical laboratories as we use in our laboratories in the research realm. How do you actually adopt and have with rapidity the ability to use some of these advanced testings that have come across in lots of the literature publications that were previously mentioned, but can actually be used to clinically alter the disease course of a patient. I say this a lot to our trainees. I'm one heat map away from cutting off my ear because this has been, everyone loves to publish their heat map. And you just stare at it like you're going to see the sailboat eventually, because someone actually, when you look at it, you say, oh, what are these differences? We are really good at publishing lots of data in a heat map. We're not good at really analyzing it. Analysis has been a huge problem for us for all of these data points. And certainly an area that we certainly need to harmonize the ability to how each of us are analyzing all of these data points in these high level research studies that we hope to move some data clinical medicine. So in Chris's study, he also talked about four different scenarios. So if you sequence an individual patient, just the patient's blood, for instance, you'll get a set of information. It's probably to the thousands. If you look at multiple organ changes, you now move to the millions. And then when you start to sequence across millions of patients, you can imagine exponentially the problem you face with all of these data points and what it all means, let alone taking more than just a single time point, but multiple time points that demonstrate a syndrome's temporality. So what are these tests that we use? I apologize. This is very simplistic to those that use these on a routine basis. I was asked if I could just mention a couple of tests that were certainly used, at least in the research lab, that could be used for personalized or precision availability. So historically, we've all looked at ELISAs and cytokine quantification. So in being able to look at both pro and hypo inflammatory markers that tell us at least upstream or downstream effects of a mechanistic change, but this is not mechanistic, it's descriptive, it's static. It's done usually in a single time point, although it can be done in a serial fashion, as you can see here in these multiple cytokines that are evaluated in this publication in sepsis. You can look at potential evolution in certain patient populations of these cytokines, and then you can actually look in cluster analyses and supervised learning what happens over time in relevance to other genomic tests. So historically, we were here, we've moved hopefully into being able to integrate different modalities to understand more approaches in what happens with patients, but this is certainly in its infancy. This is a static evaluation over time. It doesn't tell you necessarily functionality. I think one of the great things that's done, and Mark calls in the audience, is the ability to use a functional response. So using lipopolysaccharide in whole blood for four hours and then evaluate TNF alpha production, and if production is not as high as one might expect, the patients at least, cells, monocytes, are certainly not reacting where they should, and these folks are at high risk or demonstrate paralysis of their immune system and could benefit perhaps for stimulation with drugs such as GM-CSF. Dr. Hall and others are using what I think is a really exciting platform, and I know there was a mention of the smaller study with 15. Dr. Hall's study is looking at well over 1,000 patients, I think, which is really novel in our field, and this is, I think, where we're gonna be going because this can be adopted in real practice at the bedside, whereas many of the high-level omic studies that many of us are still doing in parallel are gonna be challenging until we get to a point where we can get that data in a rapid fashion as such as Dr. Hall's group is, but understanding cytokine functional response is certainly important because it tells you a little bit about functionality of specific cells. Our group has been using both ELISA as well as ELLA, which gives me seven parameters of cytokines, and also ELISPOT, which is an enzyme-linked immune absorbent spot. It's not the be-all, end-all. It's just one test also that could be used in clinical fashion where you can actually take a 96-well plate and you can stimulate specifically cells against the cytokines that you're interested in seeing production for. So for instance, we look at interfering gamma and CD3, CD28 stimulation, and so after stimulating it, one would hope to see generation of large spots or increased numbers of spots demonstrating a robustness and surrogate those cells that are activated for secreting of that cytokine of interest. For TNF, we look at LPS for IL-6 and IL-10. We certainly have other positive controls and we can look specifically in our dish at robustness and surrogate as a surrogate of what we think is a response to their both innate and adaptive immune markers. And then we can also modulate this system in the dish in real time at the same time with different drugs and also evaluate if we can boost the immune response or we can actually produce an anti-inflammatory effect. And so we did use this in COVID and this was published in 2020 in GCI Insights where we specifically looked at those red dots which are non-survivors compared to septic patients. Obviously, this was not a study that was studied at the same time. These are historical septic patients compared to patients that we had with COVID-19. This is just demonstrating proof of concept that we can demonstrate that we can measure at least those that have the highest risk for death, certainly had decreased interfering gamma production. And we further went on to look at TNF alpha production and saw a similar response that was more exaggerated than we saw in our septic cohorts. And so we use this model and we're moving this model into patients by evaluating different therapies in the dish over time and trying to understand the temporality to different conditions in specific patient populations. And we're mirroring this with mechanistic studies specifically with single cell RNA sequencing and proteomic evaluations using CyTOF and SOMA scan. And so we've got a large trial that's going on currently in adults to try to further endotype multiple patients with trauma and medical induced sepsis in the adult population. And hopefully in time can move this into pediatrics. This is just one step. And there's a number of groups that are also doing some parallel studies to this which I think are gonna be informative to at least the field so that we can descriptively hopefully apply this in real time to patients to be able to understand what drugs may be targeting specific alterations in their immune system. Now the gold standard has always been for the last 25 plus years, flow cytometry. So for functional immune phenotyping. So this measures properties of cells as they flow through the fluid suspension across an illuminated light path and allows us to look at several properties of cell populations, their size, their granularity, their relative fluorescence intensity. And it's got a number of different applications including the ability for immune phenotyping, cell sorting, understanding apoptosis, DNA content, cell cycle analysis and cell proliferation. And when you look specifically at flow cytometry and look at the identity and purity of a cell product, you can have the input as a mixture of different unlabeled and labeled cells and then be able to have your output as the number of labeled cells against the total percentage of cells. And this has been used in sepsis in a number of studies. So Guillaume Manaret has published a fair amount in not only HLADR, but in other uses of flow cytometry. And this was a nice paper in 2015 where he specifically looked at all the different flow cytometries evaluations that have been done in sepsis and then attempted to see which studies linked it with an outcome. And some of them did look at mortality and then proposed, or at least in the dish, looked at putative therapies that could be utilized in patients. And in fact, some of these studies were also coupled with some phase one and phase two trials that were done with interleukin-7 and checkpoint inhibitors. And certainly Volk in 1996, as well as Guillaume Manaret have looked at HLADR expression and activation and demonstrated that when HLA activation is diminished, this correlates with mortality and morbidity and developed an idea of certainly immune paralysis, but based off of HLADR activation or deactivation as his paper actually demonstrated. And then since his initial paper in 1997, you can see that there's a direct increase in usage of HLADR activation studies and flow cytometry in sepsis. And this has continued to increase all the way through to present time. But I think what's most interesting is that when you looked at inter-laboratory assessment of HLADR, there was pretty good agreement that across labs, this study could be used rather than just well-used and not actually being able to use across laboratories. This test at least demonstrated some success with inter-laboratory assessments. What I think is most interesting about flow though is that you use flow to demonstrate specifically looking at fluorescent intensities for the cell type of interest, but the major challenge is understanding and revealing these dim events, these populations which are hidden in the background. And so because of that, and our lab uses certainly our Cytec Aurora, we do spectral flow cytometry and multi-parametric where we use different parameters to look at different responses in our flow cytometry. We're able to get 35 different conditions rather than just looking at a single condition, looking at something such as HLADR. And so we use a 35 color panel and we're able to refine some of the populations and look at some of those things that are dim or hidden by previously just single flow cytometry based off of a single antibody, not done in a high dimensional multi-parametric fashion. So I think a lot of folks are certainly moving towards use of this to be able to further refine populations and understand different differences in sepsis. And then certainly we've moved into a world of omics of everything, genomic, transcriptomic, proteomic and metabolomic. These are all very different, but they all are done with potential ability to help enrich the population and understand expressional differences that could lead towards a better understanding about who's at risk for decline and perhaps who may respond to specific therapies. And so proteomic CyTOF and SomaScanner2 that my lab uses, and CyTOF is cytometry by time of flight. Unlike using specifically fluorescence to label and detect cells, which is used in flow cytometry, CyTOF uses mass spectrometry and labels with metal isotopes instead of these fluorophores and demonstrates that each of these isotopes as a unique mass, allows for simultaneous detection of multiple markers, and you're able to actually barcode and run over about a two day period to be able to get some meaningful data. Now, one of the challenges certainly with CyTOF is that data analysis can take a substantial amount of time to be able to make understanding of what the data that you have before you. Colleague Isaiah Turnbull during the COVID pandemic utilized CyTOF, I think pretty well and of differentiating between severe COVID and moderate COVID patients and partnered his CyTOF work with Luminex cytokine assays and ELISPOT and demonstrated certainly in his Luminex with cytokines that there certainly were differences in cytokine production across severity of illness in COVID not surprising in a lot of other labs at the time have certainly demonstrated this across these 36 plasma cytokines. But what was really interesting with CyTOF with Isaiah is that he was able to look at 75 features per sample using what he called a signaling phosphoproteome and across different neutrophils, monocytes, CD4, CD8s and NK cells, was able to decipher that in his readout across our patients with COVID and it was about 50 patients or so that STAT3 was the predominant signaling pathway that was activated in COVID-19. And why that was important was because STAT3 knockout mice certainly have defects in bacterial clearance during their sepsis and have also an exaggerated systemic inflammatory response. And at the time, many of us were hearing about the cytokine storm and our lab was demonstrating a predominant immunosuppressive phenomenon in this concept that perhaps you can have this exaggerated systemic inflammatory response while having these perturbations in STAT3 and having defects perhaps in even viral clearance started to make at least mechanistic sense. And certainly this is the same pathway that Barsitinib is used and could understand exactly why Barsitinib may have had efficacy for some patients with COVID. And then furthermore, genomics and other transcriptomic approaches look at either bulk RNA or single RNA sequencing. Transcriptomics, these are proprietary names for companies. Nanopore and nanostring are two types that you can look. This certainly looks at the abundance of multi mRNA transcripts within a biologic sample simultaneously. I'm not gonna go through the intricacies of certainly looking at single cell RNA sequencing, but the price for this in the research lab has come down, but the problem still relies is the analysis. And so as we move towards more machine learning of a lot of these data, I think we'll actually be able to accelerate the ability to use this in a meaningful fashion to look at which patients certainly have expressional differences with gain or loss of certainly function. And then when you looked at whole blood transcriptomics, this was from Yang and Critical Care just last year. In this paper, they were able to demonstrate some subclasses of pediatric septic shock. And by further refining these populations and then partnering this with assays that can be done in more real time, such as LPS stimulated whole blood for TNF alpha production, or use of ferritin or other nonspecific markers, you can start to understand how you can start to enrich the populations and then develop therapeutic structures to give the patients that might benefit from a drug and be able to demonstrate efficacy so that you're not using a one size fits all that all patients with sepsis are the same. And so we use nanostring as our transcriptomic of choice. We've got a specific, about 860 targets that we use that were developed by the late Hector Wong, great friend and colleague, who developed a really wonderful use of these 860 targets, which we've adopted into our laboratory. We can get our results back within 72 hours after we send our samples away from our laboratory for processing, but it allows for barcoding and then we have software that actually has some automated analysis plans and we can get results usually within about two weeks after looking at the data through automation. So this is what we really hope for. How do you find the right treatment for the right patient? So I previously mentioned that many of these individual studies to be able to demonstrate efficacy, because if you look across the board, there's about a 1% efficacy with many of these different drugs, but each of these studies would need about 10,000 patients to demonstrate efficacy. And that's because of the heterogeneity likely in the populations. And so some of the tests that we just discussed may further at least enrich and refine those populations so that we can get towards use of many of these drugs in a repurposed fashion that actually had pretty tolerable safety effects when they were done in phase one and phase two trials. But we're just starting in my view to climb that ladder. We certainly are doing great work right now with machine learning and artificial intelligence to describe clinical phenotypes. We're working out some of the nuances to develop the analysis plans for immune endotypes. And we're just starting, I think, to step that ladder to eventually get to use of successful biomarkers and endotypes across the globe to be able to target, I think, specific therapies with patients in sepsis. I think we'll hopefully be there in the next 10 years. I think this is the next 10 years that are probably gonna be the most important for sepsis because I think we're starting to utilize many of the research strategies that a lot of us have been using in the laboratory, hopefully can be moved to clinical practice. And sadly, although beneficial, I think COVID accelerated it for us. I think the way that we approached COVID may actually help us with our development of precision medicine tools for sepsis. And so Carolyn Kalfi and Pratik Sinha have certainly, I think, revolutionized their approach in models of precision medicine where they've tried to develop with both the EyeSpy network and similar networks use of randomized control trials that take into tools for biologic phenotyping, develop eventually clinical implementable models, and then hopefully can then utilize these to discover but also utilize these models for evaluation of targets. And so as mentioned earlier in the welcome, things like HLA-8s in sepsis and MIC and MAS and pancreatitis and trauma and a number of different syndromes that have a lots of overlapping clinical similarity can certainly be sampled for these patients in time in the intensive care unit, where we can further refine the different biological factors, understand the physiologic and biometric changes in the intensive care units, be able to take these high level, high throughput genetic and transcriptomic and proteomic evaluations, be able to use this in a machine learning fashion, be able to then further refine the physiologic states of our interest, move this eventually to a point of care, but be able to get these novel trial designs that allow for different treatments specific to true integration and harmonization across these different biomarker and omic evaluations. And I started our conversation to say, we do need to develop new biologic disease nomenclature. And this is not gonna be easy because individual patients are gonna be different across the board, but there's gonna be a lot of similarities. And so further defining the syndrome into more specific biologic conditions may be helpful for us to be able to deploy appropriate therapies. And as we do that and do this with folks that work in areas that I don't work in, implementation science and data sciences and AI and technology development, this is how we're gonna get to the appropriately designed, I think, randomized control trial and move towards adoption and practice and predict and prevent long-term outcomes for those that survive their sepsis. And so I mentioned that we need to integrate all these individuals in the research realm to the clinical practice, to be able to find ways to understand disease and the biology of the disease and further improvement of diagnosis and therapeutic applications. But how do we get there? Well, we need to try to find alignment in basic science for mechanisms. We need to start to work on what are the best biomarkers that can be used today, but how can they reflect some of those more advanced bench work laboratory studies that are currently being used and reflect them at least to some degree and develop them even further and harmonize them across institutions so we can operationalize approaches that could be used in the ICU and then furthermore in the post-ICU environment. How can we look at outcome measures that reflect those biomarkers to demonstrate efficacy? How do we integrate our data with AI and machine learning? How can we come up with better novel trial designs? For instance, tomorrow morning, great talk for platform designs and adaptive trial designs. I think it seems like the right approach when we use metrics that would make sense to be able to alter our randomized control trials and hopefully have some understanding of the biology. And how can we create larger collaborations? So this is gonna sound odd, but if you think about the Human Genome Project, a large, well-organized, highly collaborative international effort that generated the first sequence of our human genome and that of several additional well-studied. It occurred from 1990 to 2003. It was one of the most ambitious and important scientific endeavors in human history. It occurred from 11 donors. That's it. Not thousands of patients, 11 donors. And they were able to at least give us the first pass at this over that 13 years. How many people have died since the conclusion of this? 220 million people have died from sepsis since we concluded the Human Genome Project. Yet 7 million people have died, sadly, from COVID. Huge differences of scale. And so I think this is an indecent or decent proposal, but in my view, if I can pontificate for three more minutes, if we're able to pay quarterbacks and baseball players $600 million, maybe we need a sepsis project to understand the scale of this, where we can actually bring all disciplines to the table, systems biologists, AI, machine learning, trialists, immunologists, hematologists, microfluidic engineers, where we look at the same patient samples and across different patients, we engage different countries, we look at the same pathogen and different pathogens, we look at different evaluations at each site that can conduct high-level basic science of biology, but then are in a position to be able to conduct follow-up adaptive or platform trials. In my view, we need this, because I think we're going to work in a lot of silos and understanding specific areas that each of us have some expertise in, but we may not understand the true complete understanding of this syndrome and be able to move this, I think, into a better way to influence more patients, unless we do something like this. That's my two cents. And so, because of this, I did engage a conversation with folks at the NIH and at Google who actually were interested, and not something I can lead, it's not mine, but something certainly that hopefully we can develop somewhat of experts in the future, and surprisingly, the NIH was interested in holding that summit. So, could be a first step forward. Obviously, it's going to take a lot of years of planning to get agreement across a lot of different scientists, but maybe it's the right step, or maybe not, but I think perhaps it's a step that could help us move towards understanding the syndrome. And so, sepsis represents a complicated biology and a heterogeneous set of host immune, hematologic, and biologic system responses, we heard about this today, that require further characterization into usable, immune, and clinical endotypes to inform our therapy. We do need to move to personalized medicine. We do need to understand the biology of the syndrome before we create more clinical definitions and more subclasses so that we can understand and refine the heterogeneity of actually what's occurring in a functional manner, and perhaps use that to develop further functional tests. And then certainly the data explosion needs professional help, something I can't do, but those that have professional experience that may not be MDs and PhDs, at least in the medical field, but may be in the world of computer science and engineering and other areas, I think that's probably where we need some assistance to be able to take all of these wonderful pieces of data that many of us are working so arduously to collect and be able to package it in a fashion that could be acted upon. And I do think an accelerated pathway to initiate a large sepsis collaboration that brings together industry, federal funding, philanthropic donations to improve precision medicine is due. If we've lost almost a quarter of a billion people since the end of the Humans Genome Project, I hope that we could hopefully save more lives in the next 20 years with this sort of large set collaboration. I think that's how we can best understand sepsis and hopefully move the field forward.

precision medicine

sepsis management

ELISpot

flow cytometry

transcriptomics

proteomics

metabolomics

artificial intelligence