Scientist Stories: Elaine Mardis, Genomics & Frontiers in Oncology

Good morning. I'm happy to introduce the Frontiers in Oncology for July this morning. We're very pleased to have Elaine Martis with us today. Elaine is the co-executive director of the Steve and Cindy Rasmussen Institute for Genomic Medicine at Nationwide Children's Hospital in Ohio. And holds the nationwide endowed chair of genomic medicine. She also is a professor of pediatrics at The Ohio State University College of Medicine. Elaine initially graduated from University of Oklahoma with a BS in zoology and got a PhD there in chemistry and biochemistry. She then did postgraduate work in R&D at the Bio-Rad Laboratories, and then was a member of the faculty of genetics and co-director of the Genome Institute at Washington University School of Medicine from 1993 to 2016. And as you know, was very involved in the Human Genome Project there. She's gotten many awards and honors, and most recently was the immediate past president of ACR. She's a fellow of the ACR Academy and was just recently elected to membership of the National Academy of Medicine. So please welcome me, please join me in welcoming Elaine today, who's gonna give a talk on the genomics Guided Pediatric Precision Medicine. Elaine.

Thanks so much, Jim. It's a real pleasure to be here, and I'm happy to present at Stanford, although when I've done that in the past, it's always been in person, so I miss my trip out to lovely Northern California. So I, I really hope you can benefit and, and learn a bit from the work that we're doing here at Nationwide Children's Hospital around on pediatric precision medicine. I'll focus today in particular on pediatric cancer because cancer is the leading cause of death by disease in the developed world in children over 1 year of age. And of course, we've known for some time now due to a variety of pioneers in the field that cancer is a disease of the genome. Here's one such pioneer, Dr. Janet Rowley, at her microscope looking at cancer chromosomes as shown on the right-hand side of the slide. She was really one of the pioneers who, from a technological standpoint, learned to condense chromosomes and stain them so that they could be viewed under the microscope, and then started systematically working through patient samples to show that canonical translocations such as the t(15;17) here in acute promyeloacidic leukemia were reproducible from one patient to the next. And so what we use today is obviously not so much the microscope to look at chromosomes, but next-generation sequencing instruments. And what I'll talk about today is that in 2020, our Institute for Genomic Medicine used next-generation sequencing and computational analysis to study over 100 children with cancer. And we returned medically relevant information on over 90% of these cases. I'll talk to you today about some of the work that we've been doing in this regard. I want to set the stage for the discussion though in a context of pediatric cancer survivorship because this is really the challenge that we are up against in this day and age. In particular, there was a report in JAMA Oncology 2020 just last year really looking back over the decades of pediatric cancer survivorship and various aspects of mortality in those patients. On the left that received chemotherapy alone or on the right that received radiotherapy alone. Some of this work, you know, historically has led us to reduce dosage, for example, in radiotherapy, especially in younger children. But if you really look at the decades from the '70s to the '80s to the '90s, which would be the decade the child was diagnosed with cancer, you can see that it continues to be the case that survivors of pediatric cancer still experience excess causes of mortality from a variety of modalities, including pulmonary, cardiac, and subsequent cancers. And what I'll talk about today is that even though we do see long-term survival improving, we really have to commit to changing this reality, primarily through the use of precision medication, much like we've seen happen in adult cancer setting. So that's really what I'll focus on a lot today, and not necessarily in all targeted therapies, but I'll finish by focusing on immunotherapies, which is another area where we don't commonly think about children receiving immunotherapy, such as immune checkpoint blockade therapy. So just to set the stage, obviously, it is the case that there is a rationale for targeted and immunotherapy. Mike Berger and I put together this review a couple of years ago, really charting out how genomics was playing an increasing role in the adult setting with regard to both the identification of somatic and germline variants and their translation into therapeutic modalities. And so that's really the paradigm that we would like to begin embracing in the pediatric cancer setting, but it's really been very slow to develop. What are the reasons why? Well, on the left, we know that unlike adult cancers, pediatric cancers typically have very few mutations. And even if we can identify an FDA-approved drug that could target the specific mutated gene, we don't really always understand dosage or adverse event profiles for kids that are not just small adults. They have very unique, um, systems biology going on, and they have very unique immune systems as well. And so this is one of the challenges Also complicated by the fact that clinical trial enrollment historically has excluded pediatric patients in adult trials, of course, and that pediatric trials really haven't had the sufficient numbers to accrue, especially in a single setting, a sufficient number of patients to really fuel a clinical trial for reliable outcome in terms of what— how we conventionally think about clinical trials. I already mentioned checkpoint blockade immunotherapies are rarely are even considered to be appropriate, often due to the low mutation burden, as I mentioned. And so what we really need to do is to start contextualizing the pediatric cancer genome in a multitude of ways, looking at both DNA and RNA, as I'll talk about today, to really help alleviate these long-term consequences of aggressive chemo and radiation therapies when possible, and really using that paradigm to improve overall survival. And decrease morbidities. So along these lines, a few years ago now, when I came to Nationwide Children's Hospital, we introduced this cancer protocol that was at the very beginning of 2018. And we've been running it ever since. The eligibility for this research trial is for our patients with rare relapse or treatment-refractory cancers, as well as rare hematologic diseases to be studied. The profiling is outlined below. So we offer first an enhanced tumor-normal exome comparing tumor DNA to normal DNA. We call this enhanced because in addition to all of the known coding genes, the exome as it's defined, we also provide probes into enhanced coverage of cancer-associated genes to give us copy number information across all of the chromosomes. For those of you who know pediatric cancer, this can often be diagnostic or prognostic, which which can, of course, lead to differences in therapeutic approaches. We also generate relatively high coverage, 250-fold coverage depth on the tumor to really understand heterogeneity as well. In addition to DNA, we do full RNA sequencing from the tumor and do multiple analyses. I'll talk a lot about those today. We introduced a little while back the Illumina EPIC 850K methylation array. And for that, we primarily focus on CNS and increasingly now sarcoma classification because of the wonderful approach to analyzing these data offered by the DKFZ over in Heidelberg. And then finally, we have targeted ArcherDx assays that are looking at tumor RNA to identify driver fusions. So I've put stars next to 3 of these assays, the tumor-normal exome, the methylation array, and the ArcherDx assays. And we now have moved these assays from the research space where they've been extensively vetted, as I'll tell you about today, into our CAP-CLIA environment. So IGM is a unique environment that combines both research-based genomics with clinical genomics, microarray, and cytogenetics as well. So you can't do modern-day genomics without advanced computational capacity and pipelines. And so what I've outlined here are just some of the computational analyses that were— are performed on our tumor-normal exome along the top. And then at the bottom, the RNA-seq from the tumor. And what you can see, of course, if you look in the center, is that we're doing rapid alignment to the genome. Here we use our Churchill aligner, which is why I have the picture of Winston Churchill here to remind me to focus you on that. And there's a corresponding somewhat outdated publication indicated at the bottom. But using this approach with the exome, as I mentioned, we can be pretty comprehensive about what we can find— SNPs, indels, structural variants, copy number, LOH and ROH as well. And then with the RNA-Seq, we really see the add value aspect of RNA-Seq analyses where we can go through fusion detection using an approach that I'll tell you about. We're working on variant calling from RNA, but certainly if we identify variants in the DNA, we can look to see if they're expressed as a 1-to-1. And then along the bottom, we are extensively profiling these patients with respect to single sample gene set enrichment analysis for outliers, looking also at the immune component with the tumor immune microenvironment with deconvolution approaches and others, and then performing principal components-based comparisons back to large databases of pediatric cancers. This can be especially helpful in settings where there's an indeterminate diagnosis from pathology and we really don't understand what's involved with that patient or how best to diagnose and treat. One of the really valuable databases comes from just a little bit south of where you are at Santa Cruz, where Alina Vaski and David Haussler's group has pioneered the Treehouse Initiative, offering a variety of online tools that really only require a small list of the top expressed genes to go into this Treehouse database and compare your patient to several hundred other patients, including in the pediatric realm. And then we also use the treehouse analysis shown here on the left-hand schema to compare our patients into the treehouse cohort using this approach to gene expression outlier analysis, or indeed just to the individual patients in our own cohort. Since the numbers now are quite large for our patients, it's not as important as it used to be to go to the treehouse, but that really was critical to this type of analysis when we didn't have many patients enrolled in our protocol. And as you can see in the orange boxes at the bottom, one of the things that we look for in this type of outlier analysis using databases such as the Drug-Gene Interaction Database, DGIDB, or Ingenuity Pathway Analysis, is the potential for druggability for highly expressed genes. Of course, the challenge there is then going to the immunohistochemistry world to determine whether there's a pathology-based stain that can validate that overexpressed protein. But in particular, we do evaluate patients in this way. I'll just give you a quick example here of a patient that we have studied looking at this patient CH198 shown by the red dot. This is a Ewing sarcoma patient over to the rightmost aspect of this plot. We're looking at this patient's expression versus Treehouse for all other Ewing sarcoma that are in the Treehouse cohort. And then on the right-hand side, we're comparing this patient to our own patient population, showing high expression in JAK1, the gene of interest for both of these analyses. So this is just an example of a druggable outlier that we could evaluate in this Ewing sarcoma patient. I mentioned fusion detection and the importance of fusions in pediatric cancer. Indeed, many of these fusions are druggable. And we've developed this rather complex but highly effective fusion detection pipeline shown here that takes our RNA sequencing data, shown at the top on the right diagram, down through multiple different predictive algorithms that look for fusions. They all have their own strengths and weaknesses, properties that are highly complementary, because what we do is we evaluate the predicted fusions across all of these callers through an overlap script that really gives us a collective look at the commonly called fusions. We can then prioritize those with respect to any known fusions, including knowledge-based filtering. The hierarchical prioritization has to do with how many callers actually called that fusion in particular. And what we've found over many patients' worth of experience is that if we have at least 3 of these callers predicting a fusion, we have a 100% confirmation rate when we take that fusion back to our clinical laboratory for confirmation with just a directed assay. And so we, at some point in time, will CAP-CLIA validate this as our first approach into the CAP-CLIA validation of RNA sequencing. And we'll follow with other types of analyses. So just a quick example of a newer additive algorithm to this ensemble approach is Cicero, which came from St. Jude. And the Cicero algorithm has internal tandem duplication detection as well as fusion detection. Here's a beautiful example of an FGFR1 fusion— sorry, internal tandem duplication for this particular patient that was identified by Cicero. These are incredibly hard to detect with high accuracy. So we're really pleased with the addition of this algorithm. Also, this patient showing a high expression level for FGFR1. That's a druggable target. Finally, using the Cybersort X algorithm, which I think many of you at Stanford are probably familiar with, we're also deconvoluting the RNA-seq data to evaluate for immune cell infiltration. And I'll show some examples of Cybersort in my real data examples here in just a minute. And then we're also commonly— and especially in cases where patients come to us with formalin-fixed tissue only available from the tumor— utilizing the complementary NanoString immunoprofiler assay which evaluates a variety of different immune response categories but also tells us about the immune cell content based on the gene expression profiles for known immune cell types. So finally, just an area where I've been really pleased to do some work while I was at WashU and continuing on more recently with those collaborators at WashU, we also are able to predict neoantigens using our PVacTools suite of immunogenomic predictors shown here in a figure from the recent paper in Cancer Immunology Research. And we continue to update this as updates come along from that group at WashU to predict neoantigens, often in the context of enhanced immune infiltration. So I'll just wrap up this introductory part of the talk by showing you where we've got to in terms of this protocol, as I mentioned, that opened in 2018, enrolling patients over time. We now have studied 267 patients in this protocol. You can see the breakdown of different types of malignancies on the right-hand side of the slide, the distribution in ages, and then the sex distribution on the far left.. And in particular, you may be wondering about how vascular malformation and overgrowth syndrome snuck in here. Not cancer, obviously, but this has expanded out this protocol to include a group that we have here that focuses on kids with vascular malformation and overgrowth syndromes. Often these genes can be cancer genes that are not mutated in the form of a cancer, but rather in the somatic mosaic setting. And so this lends itself nicely to the comparison of affected versus unaffected tissue. In particular today, for my examples and for the research that I do in my lab, which is a spin-out of the CANCER protocol, I want to focus on CNS malignancies and in particular just show you a breakdown here. Over 50% of the patients that we've studied on this protocol have had CNS malignancies, and you can see that they predominate in astrocytic and oligodendroglial tumor tumors, rather, embryonal tumors, and so on. Looking at the bar graph that's labeled up with the different types of tumors that we've looked at. And in particular, what I'll focus on today is primary and recurrent tumors in the CNS setting. Finally, just to finish with the tale of where we've got to, I told you in the very first slide, we've returned medically meaningful results in over 90% of the cases. Studied on this protocol. Medically meaningful to me means we got diagnostic, prognostic, or therapeutically relevant information out of the combination of tumor normal exome RNA-Seq and the methylation arrays. And you can see here along the bottom, um, the various breakdowns in terms of the number of patients and the number of alterations of that type that have been identified across the spectrum from germline predisposition all the way to medically meaningful somatic variation, which often means that targetable therapy is indicated. So quite a good yield, if you will, in terms of the clinical benefit for these patients, potentially, if the therapy gets used. And that's a discussion that we can have during the, the end of the talk if, if you're interested in hearing more about that. So I wanted to wrap all of that up into just some tangible examples so that it hopefully comes across as something that's fairly coherent and reasonably meaningful, but also really perpetuates this virtuous cycle of studying patients, but also benefiting one's own research with new hypotheses that emerge from those studies of those patients. So we'll look first at this patient described in terms of the case along the left-hand side of this slide. The figure that you're looking at is basically the primary and recurrent tumors from this patient evaluated in the context of the Treehouse cohort principal components analysis. So looking specifically now at CNS malignancies, you can see that this patient, based on transcriptome profiling in the primary and recurrent setting, is close but not quite the same. You can see that also in the second bullet point along the left where the pathology diagnosis in 2015, the primary, was a mixed glioneuronal tumor, and in 2017 had a slightly different description transcription from the diagnosis high-grade neuroepithelial tumor. This was also sent to St. Jude Pathology, categorized differently. And you sort of get the feeling here that there's something different going on in the primary and recurrent tumors. The RNA analysis, importantly, from the fusion pipeline I just showed you identified a rare, at the time not well understood, EP300-BCOR fusion. So we took a look in the literature to try and understand this a little bit better. We were able at that point to find an adult TCGA brain tumor that had an EP300-B core fusion but with different breakpoints than our patient. Our patient shown at the bottom in terms of the different amounts left over from EP300 and B core in the fusion, if you will, the part that's deleted in the center, whereas this patient just had the entire coding region of B core preserved. So it's slightly different. But maybe not the first time a fusion between these two genes had been reported. Didn't really tell us much at that point in time, but we went and looked a little further at just patients with these internal tandem duplications, the one that FGFR1 that I talked about earlier. And these had actually been identified in this paper from 2016 in Cell in about— in a small number of patients with high-grade neuroepithelial tumors with B-COR ITDs. And so our hypothesis at the time is that there was some biological similarity perhaps between the EP300-B-COR fusion and these high-grade neuroepithelial B-COR cancers. The importance of reanalyzing patients over time came fresh to us then a couple of years later in 2019 in this paper from a group at Harvard and others, as well as UCSF, showing 3 and only 3 pediatric patients with EP300B core fusions and showing the biology as an indicator that these were drivers of pediatric gliomas with distinct clinical pathologic features. So this was the first point in time where we really had a context in which to evaluate this patient and what the potential drivers were for this patient. We, based on the results of this paper, then went back and looked at EGFR because that was also described as being overexpressed in the 3 patients in that paper that I just showed you. And indeed, you can see in the red bar and the red bars on the bar graph here, the primary at the far left and then the recurrent tumor somewhat in the first third. In terms of the level of EGFR expression. Now, you might wonder, well, it looks like the EGFR expression is lower in the recurrent tumor. While that may be true in terms of the absolute counts, it's diluted somewhat by the fact that this patient also in the recurrent setting has a very high level of immune cell infiltration. So we evaluated that in a number of different ways. First of all, you can see the estimate scores, which are again, at obtained from RNA-seq analysis along the top here for the primary and recurrent. And you can see a significant increase in positivity in the immune score going from the primary on the left to the recurrent on the right. And then similarly, we evaluated this patient's samples with Cybersort X and were able to identify just sort of in the center on the bottommost bar graph, which is the recurrent sample, a significant increase in the macrophage and monocyte infiltration in this patient's tumor. So these all sort of add up to the fact that in the transition from primary to recurrent disease, there's a recruitment of macrophages and monocytes into the, into the new tumor, into that recurrent tumor. And we further substantiated this by working with our pathologist to perform several immunohistochemistry stains You can see them here for CD45, heme-derived cells on the far left, a variety of other markers, and in particular CD163 on the far right showing in the primary on the lower panel, not much at all of the CD163 or other stains there, but significant increases in the recurrent tumor or our along the top panel. So then finally, what, what we did to analyze this a little further was to compare the primary and recurrent variants. On the left-hand side, you can see the variant allele frequency comparing the primary on the x-axis to the recurrent tumor on the y-axis. There are a couple of shared mutations, but they're quite different otherwise in terms of the variants that are found in these two tumors, certainly far more in the recurrent tumor. And this will be after in an interval of radiation therapy, which often we think enhances the certainly potential for inflammation, which is an aspect of immunity. And then on the far right-hand side, two graphs showing the predicted MHC binding affinities of tumor-specific peptides that are predicted from the mutations identified in these patient samples, showing on the far right that in the recurrent tumor, we have a significantly increased number of neoantigens, but possibly with macrophages setting up an immune suppressive environment, no opportunity for that patient's immune system to really respond to that increased neoantigen load. Um, just one other case that really illustrates some of the tools that we've developed here is a really interesting case, fairly rare. These are being seen now more and more with increased genomic profiling. But still a bit of an outlier. So this is a patient, you can see the patient's history and MRI scans here going from left to right. The patient had a presentation in September of '19, the tumor was resected and on post-op looked like a gross total resection. But just a month later, a follow-up that came along with some emerging behavioral and cognitive deficits showed that this tumor appeared to have recurred. There were a couple of attempts at therapy, radiation with filipirib, followed by temozolomide and radiation. Sorry. And then the follow-up in January of 2020 showed that those— neither of those therapeutic intervals had really touched this patient's cancer at all, and it was continuing to grow. So another resection was performed, and we were then able to to study the patient on our cancer protocol as I've described before. What we were determining from this is shown here. So I've already sort of given it away in the title, but this is a patient with Lynch syndrome. So PMS2 is mutated at a germline variant allele frequency, rather, on the top table of 49%. So an inherited alteration we know from studying the parents. And in addition, there was a VUS in polymerase epsilon, which can also contribute to high mutation load, but this was a variant of uncertain significance. However, in the somatic analysis, we found the second hit in PMS2, which is shown here. And this would be, you know, another deleterious mutation to that gene. And indeed, we did go through immunohistochemistry for Lynch syndrome proteins and were able to show with help of OSU pathology who offers these stains, that there was no PMS2 protein present in the patient's tumor. You can also see polymerase epsilon, another hit, as well as MSH3 mutations at different variant allele fractions, etc. And the bottom line is this patient has an ultra-mutated tumor genome with over 14,000 filter-pass variants that were detected. That means high quality, And you can see different cancer genes here that are really, you know, less important in terms of driving the cancer per se than they are the affiliated detritus of just broad mutational activity here based on the PMS2 alteration in this patient. So predictably, we can evaluate this patient using the methylation profiling analysis, which I talked about a little bit ago. This is the, um, methylation classifier, not the newest version, an older version. And here the patient just scores a little bit below where we'd like to actively call this as being correct. It's a 0.76 score. We really don't definitively, um, give a diagnosis based on this classifier unless the score is above 0.8. But I do think this may point out that for this particular version of the classifier there really weren't a lot of these ultra-mutated tumors that were being considered in their machine learning-based approach. I think that's probably been addressed in version 12, which is just now available. So to finish up for this patient, the pretty obvious therapeutic indication from 14,000 high-quality variants is the use of checkpoint blockade therapy. This patient was able to get nivolumab on therapy starting in June of last year. And to his most recent imaging about a month ago or so remains no evidence of disease with this ultra-mutated glioblastoma, hopefully in his past, but he will continue to get evaluated over time. And then to finish with the immune genomics component here, this is just a comparison of pathology and RNA-seq deconvolution where we're looking at in the bottommost IHC stain and Cybersort analysis. The primary tumor, looking at staining in particular with CD168, as I mentioned earlier, for M2 macrophages. And then in 3 different sections from the recurrent tumor, going up from the bottom, you can see the corresponding high staining in the immunohistochemistry and the corresponding increase in the prediction from Cybersort X for the RNA-seq from these different segments of the tumor of significant M2 macrophage infiltration. So a lot of the work that I talked about today, and much of which I didn't have time to, has really led us to think about the importance of the infiltration of immune cells and in terms of the setup of immune suppression in the context, especially of recurrent CNS cancers, where we probably have the most significant opportunity to intervene therapeutically for these patients once they have failed the standard of care. And this preliminary data that I talked about really led us to apply for a large project, a U54, which we were thankfully granted. Under the Cancer Moonshot from NCI. And this is in particular part of a pediatric immunotherapy drug discovery network to really sort of put together groups that are very interested in immunotherapy for kids but have different ways of going about it. So I think more shots on goal is always better. In particular for our grant, we are using preclinical models to evaluate several different approaches. They're listed here. Engineered natural killer cells, oncolytic virus-based therapies, and targeting the myeloid cell-based immunosuppression that I talked about a little bit in the context of our patients that I showed you. Finally, my part is comprehensive immunogenomics. This includes mass cytometry in collaboration with Greg Babahani, who trained at Stanford, who's at OSU Medical School now, and, and single single-cell RNA-seq as well. So I wanted to show you just a brief vignette on some of the single-cell data just to finish off the day. In particular, we've been focusing on oncolytic herpes simplex virus. I'm starting with 134 as described here, but also interested in variants of 134 that are being developed to sort of enhance the immune response to injection of 134 into the, into the tumor resection bed, which is really the form and flavor of clinical trials in patients that have been pursued with this virus in the past— adult patients, I should be clear. And here we're working in collaboration with Kevin Cassidy, who's also part of our U54 team of investigators. Their data has really shown that using this particular variant of OHSV produces an effective T-cell-based immune response and that can increase survival in a malignant glioma mouse model. So we've been doing some studies of this in the context of 3 murine models of medulloblastoma that were kindly provided by Martine Roussel at St. Jude. These are syngeneic models, and so they're immunocompetent. They grow in immunocompetent mice, and this allows us, of course, to evaluate the host immune response in viral therapy. And so you can see the 3 different models. Shown here that correspond to 3 of the 4 types of pediatric medulloblastoma from molecular subtyping. So we've been providing— we've been working with Kevin to do the single-cell RNA-seq analysis in C134-treated versus vehicle-treated mouse medulloblastomas. Here's just some— a teaser example of some data that we've generated. It's fairly early days here. On the left, the C134 treated is in blue and the vehicle treated in red. And you can see some unique clusters sort of up at the top that are coming from the virus-treated medulloblastoma. And you can see the assignment of different clusters on the right-hand side in terms of aspects that are common as well as unique to these different experimental conditions. And then just to finish this part, we also used kind of a unique reverse Cybersort approach to help us cluster these cell clusters and make sense out of what they are, which is often the hardest part in single-cell analysis, as far as I can tell. And so what you can see, I think, here at the top— although it's a little bit occluded on my screen from the menu from Zoom— but we've been able to basically assign predominant immune cell types in these different clusters from CybersortX and then layer them onto the corresponding cluster, helping us to better identify, for example, that in cluster 11, there's a preponderance of many different types of immune cells, including T regulatory, CD8 T cells, and B cells that are corresponding to the specifics of the C134 treated disease. This is the subgroup cluster that I was showing at the top of the prior screen, just to be clear, and now shown just in exclusivity in terms of really helping us to interpret what's going on with respect to the, to the virus-treated medulloblastoma. So again, early days, but I think this will be important as we go forward to help identify the impacts of these viruses what engineering-based changes to them make in terms of response-based changes in the mouse models, and then hopefully over time advancing this into the pediatric setting. Finally, I just want to acknowledge another thing that we're up to, to finish off, and this is sort of an umbrella trial that we've been putting together as a clinical trial concept with a recent— two recent recruits well, 3 recent recruits to Nationwide Children's Hospital. So along the bottom from left to right, you can see Maryam Falati, who is our new chief of neuro-oncology as of last August. So almost here a year now. She also brought Ralph Saloom in the middle there with her, also a neuro-oncologist. Both were at Cincinnati Children's, now in Columbus. And then Prajrajappa, a recruit to the Institute for Genomic Medicine. From the East Coast. He was training at Weill Cornell Medicine in neurosurgery and conducting precision oncology there in the pediatric setting as well. So putting this group together along with the existing experience that I've told you about already today, I think is going to be really exciting as we focus in on both pediatric patients with primary CNS malignancy diagnosis just for general genomic profiling in anticipation of the possibility they may not respond to the standard of care, and this would give them other options in the recurrent setting. And then on the right-hand side in the gray box, a specific trial with the Connect Consortium that Miriam leads, looking at pediatric AYA patients with recurrent high-grade gliomas. Of course, these are the kids with the worst outcomes in the disease. So So full-scale clinical genomic DNA and RNA profiling is planned. Liquid biopsy, which has not been established in pediatric patients as part of the standard of care at this time, will be pursued as well. And there we'll also be looking in the liquid biopsy materials for evidence of immune recruitment using a variety of approaches, which I'm happy to talk about. So we think that this is the future. This is the way to really address the high mortality from CNS cancers, but also really to improve the quality of life over time for those patients that are fortunate to survive their cancer diagnosis. So I'll finish here and just say that hopefully I've been able to illustrate that these clinical genomic assays of pediatric cancers can really add important clinical evidence. They can also provide us powerful information about cancer drivers, gene expression, but also the immune microenvironment. And this is really driving new hypothesis-driven research into the CNS tumor microenvironment, including aspects of immune suppression and new therapeutic approaches for kids. We think that it's quite clear scaling this amount of molecular profiling will be our next focus. This will include the Umbrella trial that I told you about just a minute ago, but also offering these CAP-CLIA-validated assays of DNA and RNA to our partner hospitals and also to other hospitals across the state of Ohio and beyond. We're also very focused, and I didn't have time to talk about it today, on one of the key aspects of scaling all of this from a clinical standpoint, which is automating variant interpretation using machine learning-based approaches. So we're early on in that effort, but we're pretty excited about the possibilities there for scalability, and we're gonna need it. So I'll just finish by acknowledging the whole group. It does take a village, not only within IGM as listed in the two columns on the left part of the slide, but just a significant body of collaborators across the hospital. I've just talked about cancer today, but we have other translational protocols with other groups of providers in our hospital, which saw over 2 million patient visits last year. So one of the largest in the United States. And I'll finish by just saying a special thanks to our patients and families. Really couldn't do this work without their courage and their willingness to participate. Thanks very much. And I think Jim will start forking over some questions my way.

Elaine, thanks so much for a really beautiful talk and great examples of modern precision medicine. I would just remind folks in the, in the audience, questions should be put into the Q&A and I will field them from there. So Tanya Gruber writes already, and I don't know, Elaine, if you can see these, but I'll read— I'll summarize quickly. She's kind of questioning about the turnaround time of the various assays in your pipeline and cites St. Jude's and Sloan Kettering as having, you know, pretty rapid 2-week RNA-seq turnaround time but a lot longer for the exomes. And, you know, where you guys are and how ultimately, you know, overall will decrease those times.

Yeah, thanks, Tonya. It is indeed the challenge to get these not only clinically validated, but with a TAT that, you know, matches our expectation from our providers. And of course, importantly, is clinically relevant to these patients. And I would say that we are working towards that 2-week turnaround time, 14 days from receipt of sample to output. I think that there are two challenges in that aspect. One I already mentioned, which is variant interpretation, especially when you're looking at something as broad as the exome. And, and I, you know, sort of briefly touched on how we'll try to not just hire more people to analyze genomes, because I don't think that's really the notion of scalability that we're all about here, but actually to develop out more machine learning-based approaches to inform, you know, where we pay attention. I think the other aspect, which I didn't talk about at all, and especially when you're looking at the fact that you have 3 sources of information, the tumor-normal comparison, so germline as well as somatic, the fusion analysis, which can often provide definitive drivers but sometimes doesn't, and the methylation-based classification that somehow we have to find a way in that 14-day time period to sort of knit everybody together. And that's easier for some diagnoses than it is for others. So it really scales like in adult cancers to the fact that the more frequent the diagnosis. Medulloblastoma is a great example where we know a lot about how to subtype and then subgroups of subtypes using a combinatorial of different assays. That's certainly not the case for more rare tumor types. So I think our biggest challenge to our providers is to really present them, you know, a complete but integrated data set that they can and try to derive some meaning from for those patients. And that's really going to take a Herculean effort that I think will transcend our own measly efforts. But we're really hoping and actively pursuing collaborations with other groups. So we've had some initial early discussions with the Australian Zero Childhood Cancer Initiative. I don't know if you're aware of that, but In Australia, basically, as of this year, their government has funded an initiative where every child diagnosed in Australia with cancer will have full molecular profiling offered to them. And, you know, so they're obviously databasing a lot of information along the lines of the assays that I talked about today, and we'll find, I think, increasing capabilities to get those data. They are openly available through EGA and to database them and to continue to, you know, learn from them interactively. But I'm also excited about these international clinical trial collaborations as well, which I mentioned in the context of CONNECT. It's certainly not the only clinical trial consortium in pediatric patients that is combining the knowledge in countries like Australia, countries like Germany, Great Britain, Canada, and ultimately the US. They're kind of outpacing us a bit already, primarily due to socialized medicine, I predict. But I think, you know, we're going to be making great strides to try and pick that up. So to answer your question, we're probably somewhere in the 3- to 4-week timeframe right now to get all of that information bundled up and pushed back to the provider. We have been able to expedite cases in the 2-week time frame, so we know we can do it. But I think the scalability of that is really going to be our biggest challenge.

Yeah, that's great. Um, you know, another challenge that you mentioned, particularly pediatrics, is, um, is just access to some of the drugs. You know, as you said, you know, there are variants, particularly some of the fusions, very specific to pediatrics So I wonder, do you feel that, you know, pharma is developing targeted agents to some of these that are so specific to peds, or is that kind of untouched so far?

Yeah, no, it's a good question, Jim. I think they've been incentivized as much as they can be. So, um, you'll, you'll, you'll be familiar with the Race for Children Act, which was again renewed by Congress just a few months ago. Um, I know and talk with Nancy Goodman frequently, and as do many of us, and And so for people who aren't aware, Race for Children really is an act that incentivizes pharma companies to open clinical trial arms specifically for pediatric patients. And the incentivization aspect there has to do with basically chits that they can use to sort of fast-track drugs through the FDA, if you will, obviously with the appropriate data to support the drug efficacy, et cetera. So that's one way to do it. Also a result of legislation has allowed pharma companies to drop the age, the youngest age of entry into clinical trials to, I believe, 12 years of age. So we're sort of skirting the edge there of AYA pediatric. But it's still really, I think, a dearth of pharmaceutical interest, if you will, in terms of things that are probably unique to pediatrics in particular that are still going to be left untouched. And I guess to some extent, I'm also always hopeful that the You know, academic drug development enterprise may step in in some regards here. Obviously, you know, again, it's not as financially lucrative potentially, but, um, but may, you know, provide some innovation around pediatric-specific approaches. And, and I guess finally, you know, in with my immunogenomics hat on, I also think that there are a variety of, for example, fusions that at the junction protein may be highly immunogenic and may provide a sort of mechanism by which these specific drivers can be recognized using, you know, engineered cells of different types as well. So CAR-T, for example, is one area where people are actively looking now. Even for common driver fusions, that could be a home run because you're really looking at not a precision approach, if you will, to one patient, but a broadly experienced fusion that just happens to be immunogenic, of course, in the right HLA haplotype, et cetera. So I think the more we know about the genomics, the more the possibilities expand, if you will. But it's going to continue to be a hard fight, I think, in terms of therapeutic modalities that are really going to be appropriate and applicable to children. So we'll keep pushing. Yeah.

Hey, Christina Curtis asks, can you speak to how you see the spatially resolved multiplex assays integrating with the genomics pipelines and the utility for exploring immune phenotypes? It seems there is an immediate utility for validation as the database grows for discovery efforts.

Yeah, thanks, Christina. It's a great question, and I didn't include anything in there of our recent genomic geomics work, rather. But absolutely agree that this is one— it wasn't a modality that was available when we submitted the U54, but we are increasingly performing the spatial profiling using the NanoString Geomics platform, just one of many that are available. Obviously not the one that came out of Stanford, sorry, but But I think you're right, this is going to be incredibly valuable for positioning the before and after of therapy in terms of immune cell infiltration. You know, are we enhancing it? Are we not doing anything to get those T cells in or those NK cells in, for example? So we're really excited about that and also excited about overlaying the sort of antibody-based you know, spatial profiling with the RNA expression profiling, because I think that will give us a much more complete picture of what those immune cells are doing from the standpoint of the biology that's happening there in situ. You know, again, maybe even in the before and after therapy intermediate. So yeah, a great set of tools. I think everybody's really cozying up to those in terms of the increased information about the spatial environment. And I guess another question there would be, is that a research tool or is it a clinical tool? Clearly it's a research tool for now, although I can't help but think about the future possibilities for that becoming a clinical aspect of diagnosis. That we could, you know, potentially use moving forward. So we're always thinking about that, especially since immunohistochemistry, which is really what that is, is, you know, so well established in pathology right now today, obviously with less multidimensionality and, and, you know, different, you know, detection, granted. But I think there are clear parallels though there that are going to make this a very important diagnostic tool in the future.

Well, great. I don't see any other questions in the Q&A, so I guess we'll wrap things up. Elaine, thanks so much. That was a really elegant and informative talk, and thanks so much for joining us.

We appreciate it. Yes, appreciate everybody's attending, and have a, have a nice summer. Okay, thank you. All right, thanks, Jim.