All right, good afternoon, everybody, and welcome. My name's Craig, great to meet all of you. Like we said, the given title for this talk is the Recipe for Translational and Genomic Target Therapy in ARDS Subphenotypes, and I'm going to be speaking to that. No relevant financial conflicts to disclose here. We will be talking about some off-label use of both medications and medical devices. You just heard from two world experts in this subject, Drs. Kalfi and Wilson. I'm a practicing clinician who takes care of patients with ARDS. If the audience will indulge me in a quick show of hands, practicing clinicians who are not experts in integrative omics, hands up. Okay, great, I find myself in really good company here. So I'm gonna offer some perspectives on integrative omics from my kind of viewpoint, my standpoint as a practicing clinician, and I hope that it will be illustrative for all of you, both to understand kind of where we are right now, some of the work that's already been done, and where we might be going in the future. I'll also just let you know that I'm gonna use humor to try to diffuse some of the imposter syndrome that I feel right now. There's gonna be some bad metaphors, there's gonna be some bad jokes. I have an iceberg picture, actually two of them, so just a heads up. So we're gonna summarize some key omics concepts. I'm gonna talk about both opportunities and barriers to their translation, and we're out the bedside treating patients with ARDS. Like I said, we're gonna talk about some forward-looking things, and I'm gonna highlight just a very, very, very narrow sliver of the relevant peer-reviewed literature, because otherwise it becomes overwhelming. So some of you in the audience are probably wondering right now, omics. What does that mean? Hands up, don't be shy. Omics, if you have no idea what I'm talking about, okay, a few hands have gone up, so I'm gonna speak to that now. So like we just talked about, ARDS is absolutely heterogeneous clinical syndrome, and we've identified sub-phenotypes. I'm not gonna beat that horse anymore. And we've created a hyper and hypo-inflammatory construct. It's not just a conceptual construct, it's an actual construct, a biologic and physiologic construct that we encounter every day when we treat patients with ARDS. But as we'll explore in a little bit, it's kind of just right below the surface of what's evident to us sometimes as clinicians. And having said that, not all causes of heterogeneity in ARDS are readily apparent to us, using either the physiologic or the clinical data that we have, or the biologic data that are available to us. And this, to me, is really where omics and integrative omics kind of comes into play. So this is my first iceberg, and I'm so sorry. So up here above the water, you have the kind of apparent causes of heterogeneity. We've talked about some of these things. You see these day to day in your clinical practice. The P to F ratio that varies, the patient's compliance that varies. We see these things, these are familiar to us. We have kind of lurking under the surface these occult causes of heterogeneity that are not as apparent to us when we're at the bedside. So kind of right at the surface of the water, we have these latent classes. These are, as we just saw, latent class analysis has used routinely recorded clinical data to help shed light on causes of heterogeneity in ARDS. So maybe we needed some binoculars, if I extend the metaphor, to help understand this. We talked about how we've used machine learning approaches to help see this. Again, with data that was already there, data that we interact with on a routine basis, but maybe we had overlooked this latent, kind of latent heterogeneity. Well, deeper under the water, what do we have? We have biomarkers, like we just talked about. We can't assess these routinely in our clinical practice, at least most of us can't. Down here, we have omics, way down deep. And if you'll allow me one more, it's like you need a submarine to get down there. So that's what we're doing right now. We're going down to the depths. So if it's been a while since your bio or your biochem, what are we talking about? So again, DNA undergoes transcription, becomes RNA. RNA undergoes translation, becomes proteins. Proteins and a bunch of other stuff gets metabolized, becomes metabolites. And these are the omics, right? Omics, transcriptomics, proteomics, metabolomics. So these are the omics in a nutshell. So if you came in here having no idea what I was talking about with omics, this is it. These are the omics that we're talking about. These kind of four very broad disciplines of study, but also as we'll explore, each have an enormous amount of depth to them as well. And I'm gonna make you an expert on all four in the next 10 minutes, I'm kidding, it's impossible. Okay, so moving on. So let's contextualize the omics in ARDS here real quick. So when we think about genomics, you might imagine that in examining the genome, and I'm gonna talk a little bit more about this in a little bit, you can look at the association, and we have examined the association between genetic variants and ARDS susceptibility and outcomes in ARDS. And we might be able to use genetics to identify targets that we could modulate to prevent or treat ARDS in the future. If we take kind of one step down further down the line and take a look at the transcriptome, and we've again talked about this already a little bit, again, there might be prognostic value there, we might be able to identify targets that would be suitable for therapeutic intervention. One more thing down the line to proteomics, so a lot of our work with biomarkers, again, kind of broadly falls into this category of proteomics, you can do kind of untargeted discovery where you're taking a look at the proteome in general and seeing what's up, or if you've identified some biomarkers that you think are interesting, you can do more targeted evaluation of those in plasma or in lung samples, like we were talking about respiratory samples. I'm gonna go back. And again, you might find some value there. I'm not gonna spend a lot of time talking about metabolomics because it's a more emerging field of study compared to these other three categories. But again, I think you'll hear more about that in particular over the coming years. So like we talked about, in order for these to be useful to us as bedside clinicians, and I'm trying to anchor to that because like we just saw, many of us are practicing at the bedside taking care of patients at the ARDS and not experts in integrative omics, what are the prerequisites that we need? These, we need to be clinically accessible. The information's gotta be accessible to us and it's gotta be actionable, which is to say that again, patients should respond differently to treatment. And we're gonna talk about whether or not omics is gonna pass muster in these two important dimensions. So right now, I wanna walk through an example of genomics and integrative omics. And when we think about genomics, again, the premise here is that genes influence our host response. And host responses, as we spend a lot of time talking about, are heterogeneous. So it's possible then that genomics might facilitate a better understanding of individual clinical risk and some of these drivers of biologic and physiologic heterogeneity. Again, this is the kind of premise. And you may be asking yourself right now, is this purely conceptual? I don't know if any of you are wondering, is this a conceptual construct? What work have we done in this field? Well, for those of us who may open the journal and see something that says genetics or genomics in the title, or heaven forbid, it's got like a locus in there and you say, and you flip past it, I'm gonna talk a little bit about kind of the work that's been done in genomics and ARDS to date for just a moment, because I think it's very interesting. And I'm sorry about the slides randomly changing themselves. It's adding an element of danger, which is fun. So over here on the left-hand side of the screen, we have a whole bunch of different physiologic mechanisms that have been implicated in either the development or the progression of ARDS. So you see things like vascular permeability, fibrosis, coagulation, things that we all know are implicated in ARDS. And over here on the side, you see just a handful of genes that have been implicated in this. And this list, it just goes on and on and on and on and on. And it's really quite mind-blowing, actually, I think. So we've made a lot of progress in this field already. In fact, it's so much progress that it's very, very difficult to keep up with. And I think that for, again, for those of us who are clinicians at the bedside treating these patients, the bottom line here that we need to focus on is that there's been a lot of work going on in this genomic space, but we have a lot of opportunities still for translation. So I'm going to talk a little bit about, hopefully, how we're going to get there. So let's imagine, for example, and again, this is not conceptual, we've done it, I'm going to walk through an investigation in just a second, that you do whole genome sequencing on patients with and without ARDS. And after you do whole genome sequencing, you take a look at chromosomal variants, and their association with your trait of interest, let's say here it's ARDS. You take a look at all these different genetic variants, you plot that with a scatter plot, it's called the Manhattan plot, it kind of looks like buildings, and you identify chromosomal variants that are associated with the trait that you're interested in. You then do something called linkage disequilibrium analysis which, broadly speaking, is looking for, usually single nucleotide polymorphisms that are particularly associated with the trait that you're interested in. And this, broadly speaking, is how you do what's called a genome-wide association study, or GWAS. So the next time that you're thumbing through the journal, and you see something with GWAS in the title, genome-wide association study, this is what's happening. And genome-wide association studies are fascinating, and they typically fall into a couple of different categories. So rare to see findings from genome-wide association studies with common variants with large effects, we just don't see that. You know, more commonly, what we see are, in this kind of like middle band here, studies where we see less common variants with moderate effects. And I've put some examples up here that are probably familiar to you, again, from the literature in this space, unrelated to ARDS. A lot of them fall into this kind of middle category. Some of that probably has to do with how we're wired as humans, some of that probably has to do with our analytic approaches in this space. I wanna highlight a couple of things. The first is the importance of meta-analysis here. So just like other types of investigations, meta-analysis helps us to tease out signal from noise. There can be a lot of noise in this space from time to time. Also, genome-wide association studies are just the first step. So once you identify, again, an association between genetic variation and a clinical trait or an outcome of interest, that would be of interest to us as bedside clinicians, you need to do some functional characterization, and you need to do some experimental validation. And again, you're probably wondering to yourself, what does that actually look like? What do you do when we do those things? And I'm gonna speak to that now. So I'm gonna walk through an example from Dew and colleagues that was published in Intensive Care Medicine last year, if memory serves. Again, one of many different investigations in this field, but an example of integrative omics, tying together genomics and transcriptomics. And I'm just gonna walk you through it. And again, the idea here is not necessarily to highlight the findings of this particular investigation, although I think they're very interesting. The intent here is really to help you be able to interpret and understand how we do this. So again, they did a genome-wide association study of ARDS across individuals of diverse ancestry. And again, what they were looking for here were loci that might explain some of the variation at the bedside, right? So they looked at all-cause ARDS cases, looking at patients who were at risk for developing ARDS. So again, genome-wide association study, whole genome sequencing, looking at chromosomal variance, and from there to linkage disequilibrium mapping. And they identified a single nucleotide polymorphism of interest. So once you identify a single nucleotide polymorphism of interest, the question is, what genes are implicated there? And there are a few different ways to look at that. But they identified two different candidate genes that they thought could be implicated in variation and susceptibility to ARDS in the patient population that they examined, which was using kind of a large dataset. What they found was that the expression of this particular gene, or these two candidate genes, was dysregulated by the single nucleotide polymorphism. So once you find an SNP that you think might be a culprit, you've got to go and see if manipulating that, right, manipulating that polymorphism leads to differential genetic expression. This is one of the ways of kind of validating your findings from a big data genome study. So we spent a lot of time talking about biomarkers. And what they found here was that when the single nucleotide polymorphism varied, that they didn't see differences in a lot of the biomarkers that we've talked about, IL-6, IL-8, tumor necrosis factor receptor one, all of these things, which is interesting, right? We've talked about how there's a lot of biologic variability in ARDS. So the way that I look at this is that this is kind of like another layer of the onion, right, and as we peel it back, we see causes of heterogeneity that are not readily apparent to us physiologically or maybe necessarily biologically in terms of biomarkers. Again, we're going deeper down. They also, using large publicly available datasets, found that the variation in the candidate genes was also associated with other forms of respiratory disease, not just ARDS susceptibility, things like emphysema, asthma, that sort of stuff. Then they started to validate this experimentally. So they used both a mouse and a human lung cell model of ARDS, lipopolysaccharide-induced lung injury model, and what they found here was the expression of the two candidate genes in both of these experimental models varied over time. So when they injured the lung, again, with the lipopolysaccharide model, expression of these genes changes over time, kind of peaks around four hours, goes down after there. Again, trying to demonstrate that we have a locus that's associated with a trait of interest. We've mapped that locus to two candidate genes. We're now going to do an experimental model where we injure the lung and see if expression of those genes changes. So this is that functional characterization step in an integrative omics investigation. Then what they did was they took a look at the transcriptome, and this is complicated, but what they did was they took a look at the two candidate genes and their association with the expression of a lot of different immune cells, and they found a positive association with some, like T cells, macrophages, neutrophils. They found a negative association with others, again, in humans. So again, looking here at the association between the genomic findings and the transcriptomic findings, again, integrative omics. So then the question becomes, okay, is this applicable for us at the bedside? And of course, what's been applicable for us at the bedside for a while now is COVID. So what they did was they took some, again, using an in-house data set, is they found some patients with COVID, and they were exploring the association, again, between these two candidate genes and outcomes in patients with COVID. And what they found here in dendritic cells from upper respiratory tract and also peripheral blood mononuclear cells was that relative to COVID severity, the expression of these two genes varied. So this suggests there could be some common biologic mechanism in these two candidate genes downstream, right, to explain both variability in COVID severity and susceptibility to ARDS. So in a nutshell, this is integrative omics, right? And this is a great example of what integrative omics has done for us and will do for us in the future. So again, here they used a lot of publicly available data, some in-house experimental data, used genomics to identify a locus of interest. They did functional characterization. They did experimental validation. So this is integrative omics. This is what it means, this is what it looks like, and this is what one of our future avenues of study will look like in ARDS. Some challenges for us. You know, ARDS is a consequence of other pathologies, you know, as we know, and it's also probably under-recognized. There's no specific diagnostic test for us, you know, and that can lead to misclassification bias, especially when we're using routinely recorded data. And as we just heard about from Dr. Wilson, you know, plasma samples may or may not necessarily, you know, reflect expression patterns in other areas of the body that we care about, like in particular the lung. And then, you know, since ARDS is a consequence of other pathologies, there are probably complex epigenetic influences at play here that can be difficult to conceptualize and account for. So again, there are some, going to be some substantial challenges to translation of integrative omics, you know, for us as clinicians at the bedside. So we're back here to the iceberg. So here, genome-wide association studies to date, right here above the water. What have we done? We just kind of talked about this a little bit, but a lot of them have used unselected ARDS phenotypes, right, so again, there are subphenotypes in ARDS. They may or may not respond differently to treatment. This is important for us to understand, but a lot of the genome-wide association studies have used unselected ARDS phenotypes. It's just, you know, practically, it's easier. We haven't seen a lot of genome-wide association studies reflective of a lot of diversity, diversity in terms of, you know, ancestral lineage, diversity in terms of geography, diversity in terms of race, not in the way that represents the types of patients that we take care of. So in the future, what might we want to see? Larger samples, more diverse study populations, more diversity in our references, and some more sophisticated genetic modeling. So let's say we take all of this, again, the given title for this talk, right, was the recipe for integrative therapy. So let's say we take all of the omics, we have all of the omics here, and we throw all of the omics in a pot. Is that the recipe? No, that's not the recipe, right? So what do we need to do? So I'm gonna tie together some ideas that you heard from Dr. Kalfi and some ideas that you heard from Dr. Wilson. So again, how are we gonna get this done practically as a discipline? So first is gonna be harmonized data at scale. You've probably heard a lot about harmonized big data during this course of the Congress. So I think that's gonna be step one. Step two, you know, what some people might call reverse, you know, kind of reverse discovery. So, you know, once we identify and explore some potential mechanistic pathways, that's going to inform potential candidate interventions. And we're going to need contemporary clinical trial approaches to trial these candidate interventions at scale. And I think that that's gonna be informed in large part, both in the conduct of those investigations and for us at the bedside later on, by point-of-care assays, right? So right now the problem with a lot of these things, especially biomarkers, and we talked about integrative omics, is that data's not accessible to us at the bedside, right? You know, you've heard about IL-6 and IL-8, and if you're lucky at your institution, you may have the availability to get an IL-6 level, usually after a couple of hours, sometimes after a couple of days. But, you know, you heard from Dr. Wilson about the PHIND study in passing, where they're using a point-of-care assay to get IL-6 insoluble tumor necrosis factor receptor one right there at the bedside at the point-of-care. So some of these things that seem far away are closer than we might otherwise appreciate. And for me, as a clinician who takes care of these folks at the bedside, I think that's very exciting. So again, in conclusion, integrative omics, super promising, some of it far away in terms of translation, some of it closer than we would probably appreciate. And again, I think that we have opportunities in ARDS. We recognize it's a heterogeneous phenomenon. We've made a lot of progress already. I think integrating all of that knowledge and translating all of that knowledge will have some challenges related specifically to ARDS, but we're gonna get there. And with that, I'll thank you all.

integrative omics

Acute Respiratory Distress Syndrome

heterogeneous syndrome

therapeutic targets

genomics

transcriptomics