 Okay, good afternoon. Thank you for the opportunity to speak here today. My name is Aria Kimi. I'm a kidney surgeon at Sloan Kettering. I work with James Shea, who's a medical oncologist, MD-PhD, and we study kidney cancer. And I'm going to talk about some of our work that we've done in kidney cancer post-hoc after the paper came out, essentially. So we formed what's called what we call a clinical TCGA, and this is essentially focused on questions that matter to clinicians using genomic data. So why does this matter and why is this relevant and why should we all be thinking about this? So clinical information collected at the time of TCGA is often very limited, as we just heard in terms of gastric cancer, but in many cancer types it's the same way. The data is often not reviewed in advance by disease experts focusing on the nuanced questions that matter to that specific cancer type. Cancer-specific information is often not collected. In the case of kidney cancer, only about 70 percent of the patients died from disease, and that matters in terms of when you're predicting biomarkers, whether a patient died from cancer or not. And other factors such as risk factors, post-treatment surgical treatment information, and detailed metastatic information is also lacking. So to address this, we went back to the sore sites, about 80 percent in the case of clear cell renal cell carcinoma, and collected data from disease experts. These are medical oncologists and surgeons and pathologists from the different institutions that we worked with. And in the case of Sloan-Kettering, where I work, we have a very close collaboration with Chris Sanders and the computational biology group there, and Anders Jacobson in particular helped with some of these projects that I'm going to talk about. But essentially we collected additional information on about 82 percent of TCGA, and this took about a year to collect all this data, and we also got detailed treatment information, which I'll very briefly talk about. So what kind of data did we collect? We looked at some data that was already collected, but we actually drilled down and found that there were a lot of holes in that data, such as prior cancer history, family history. We looked at comorbidities. I'm going to talk a little bit about BMI in a second. We also collected certain lab values that were lacking. We also collected whether the patients were symptomatic in presentation, which matters in particular for kidney cancer. And then in terms of metastatic disease, which I'll talk about toward the end of my talk, we looked at whether they were present at time of surgery, the location, and the timing of metastases, which matters a tremendous amount in my disease, in kidney cancer, because patients with early metastatic disease do much more poorly than those that develop solitary sites of metastases later on. And then obviously we collected systemic therapy in terms of the timing, indications, and the responses. So these were the kinds of data that we collected additional. So we collected stuff like BMI, which was not there, obviously, smoking status, systematic treatment, whether they received treatment preoperatively, which should not have happened, although 1.5% actually received treatment beforehand, and other factors that I think will be relevant down the road. So I want to talk two stories quickly in this talk, and I'm going to show you how you can use this TCJ type of data to answer specific questions. So the first part where we talk about risk factors for developing disease, and the second part where we talk about metastatic disease. So what do we know in terms of kidney cancer, in terms of epidemiologic risk factors? There are established risk factors, such as cigarette smoking, body weight, and hypertension. Now these are relatively mild associations, nothing like smoking and bladder or lung cancer. However, they're definitely well-established and well-known risk factors. The interesting thing about BMI in particular is that this is a forest plot showing meta-analysis that, while BMI is known to be a risk factor for developing kidney cancer, it has also been shown to be protective. So the patients that have higher body mass index at the time of surgery have better cancer-specific survival outcomes. So in order to answer this question, we first looked at our surgical database of about 2,000 patients, and we wanted to control for obvious confounding factors that could potentially explain why patients that are heavier do better. So we looked at factors like symptoms of presentation, whether these patients had other comorbidities that might bring them to the doctor and get screened earlier. We looked at tumor size, which could be a factor. So if you would be screened early, you might have smaller tumors. And then in order to kind of come up with a plausible hypothesis for why we see this, we then looked at about 126 patients from the earliest CTCJ. Now we have 340, but at the time we did this, we just focused to the Sloan Kettering ones, which was 126 patients. So these are just a big table, but basically the red box outlines the fact that patients that are more obese are more likely to have lower stage tumors and lower grade tumors. And when you control for factors and multivariate analysis, you see that these, that higher body mass index is associated with better outcomes. When you control for stage and grade, this does go away. Statistically, all of the hazard ratios are in the effect that you would expect in the relative relationship that you would expect. Now, and again, this is not surprising because I've shown you that BMI is so strongly correlated with lower stage disease. And this is looking at poor nutritional status. So even, so patients that have, that are more obese, they might have poor nutrition. So we controlled for levels of albumin in the serum, and we showed that this BMI is an independent predictor. So I think hopefully I've convinced you that BMI itself is associated with more favorable disease risk. So in order to kind of exploit us from a genomic standpoint, we looked at 126 patients, and we looked at, because this is the TCJ, we had a wealth of information. We looked at mutations, copy number events, promoter methylation and mRNA expression, which is where the money was for this analysis. And we performed pathway analysis of genes differentially expressed in the obese versus the normal weight cohorts. And then we just recently published this in JNCI. These are some of the figures from this, from this paper. We showed essentially that there were no differences in terms of overall mutation status, the number of mutations, including non-silent versus non-silent, that we looked at the top 15 or so genes that were mutated, there was really no difference in any of the BMI cohorts. And the same thing for copy number analysis and methylation. Essentially, no differences between the different cohorts that would have any insight into the differences in survival based on that. However, when we did look at mRNA expression, we saw some pretty interesting findings. So this is the top ranked genes in terms of over expression and under expression using log rank. And we immediately see some genes that were in the fatty acid metabolism and beta oxidation rich pathway. So ranked 8 and 12 of these were fatty acid synthesis genes, fatty acid beta oxidation genes, fatty acid metabolism process. So we clearly see immediately that there's a signal related to fatty acid synthesis. And actually, looking at the down-regulated genes, we found something very striking that more obese patients had down regulations of FASN. And FASN is one of the highlighted metabolic pathways. And kidney cancer is one of the primary figures in the nature paper showing that lower FASN levels are associated with poor outcome. And indeed, if you would stratify patients, both our cohort and then the entire remaining TCGA cohort by FASN levels, you could see that they had strikingly different levels of survival. And we know that FASN's role in neoplastic lipogenesis is well characterized. It allows for de novo lipid synthesis and essentially allows the cells to hijack the normal endogenous lipid metabolism, which you usually get from diet to promote cell survival. So FASN has been known to be upregulated in RCC before. And indeed that lower expression of FASN among obese and colorectal patients was seen in other cancer types. So essentially, the model that we have here now is that higher levels of FASN, upregulation of FASN, essentially is associated with better survival, sorry, worse survival, and the obese patients down-regulate this pathway. And we also indeed saw the up-regulation and down-regulation of ACC and both in the mRNA and protein levels. So essentially, we created a model that looked at genomic data coupled with epidemiologic data to come up with a plausible mechanism for why we see this. And it's not surprising entirely that we saw the effects in the mRNA expression, which would more postulate that there's an interaction between obesity and the milieu of the tumor as opposed to an inherent different biology within these tumors. So the last part, I would just briefly like to talk about some of our prelim data in terms of metastatic disease. Just really highlighting what we've been collecting and just our initial first pass at some of the analysis, which is just really in its infancy. So what did we collect here? So we looked at both the number and the timing of metastatic cases. So we did find that about 123 patients within the TCGA cohort had metastatic disease. But as you can see that this timing of metastatic disease was variable. So 75% of patients presented with metastatic disease. So that means that they had evidence of metastatic disease at the time of surgery. So the samples that were submitted were in the setting of metastatic disease already. Another few patients developed within the first three months. And we considered that a presentation because they were probably missed initially. And then within the first year about 17 developed metastatic disease. And then as you can see, there's a whole long list of patients that develop disease at a later point of time. We also collected sites of metastases. So very interestingly in kidney cancer, unfortunately some tumors will spread to the brain. And this is a very troubling sign when this happens that patients are usually in very bad shape. But we often don't know and standard MRIs of the brain are not what we do typically for patients unless they're having symptoms. So the questions are whether we can start seeing signals of this in the primary tumor setting. And again, the purpose of this is really just to show the data we've collected so far and talk about what we're going to be doing downstream. In terms of treatment information, you could see that we this took us quite a bit of time, but we had medical oncology fellows or physicians at different hospitals getting detailed treatment data on these patients. So we have all this information now and we're just starting to go through this data to see if we can correlate treatment responses with underlying genomic alterations in the tumors. And then we started to look at differences now focusing on high-risk tumors. So this is an analysis that Anders did. Again, this is its first pass, but we've looked at stage three versus stage four. So why are some advanced stage tumors metastasized and some don't? And now we have this information, we're taking clinical information, we're taking genomic information, and we're looking at what separates out these tumors. They both have high-risk features, yet some metastasized and some don't. And these are kind of the types of analysis that we're seeing with this. And then we're also focusing on what separates out the tumors themselves within the metastatic cohort. So can't do tumors that metastasize early versus ones that metastasize later have different underlying profiles. And just looking at PCA plus, we start seeing some differences by methylation alone, although this is, again, this is just very preliminary data. So the, and obviously new algorithms that are out there, like what Roelle has put out from M.D. Anderson, looking at the ways of deconvoluting tumors, I think this will be a very intriguing data set to look at this. As we know that certain tumors have inherent immune responses that are detectable using RNA-seq methods. So potentially we can look at the impact of the immune response on early versus later metastatic diseases, perhaps using that as a signature to determine whether a patient should be put on a different surveillance protocol or adjuvant trial based on the underlying features within that tumor. So in conclusions, the goal of the CTCGA is, can really provide powerful insights into both clinical and epidemiologic phenomenon that are very relevant. The rich genomic information can serve as discovery cohorts for targeted validations in much larger clinical cohorts, which is obviously what you need to answer these types of questions. But I think if we can really maximize the value of the TCGA, we can start answering these questions. And I think this work also highlights collaborative structures that are critical to make this kind of significant advances. Thank you very much. So at first part, you highlight a model between BMI and the molecular model. I just wonder, you don't mind which one is the cause, which one is the effect. If the BMI or its molecular change, the causal effect effect. Right, no, that's an excellent question. And I think that's remains to be seen. I would probably argue that BMI is the cause in this case, but I think those are kind of the models that we're trying to create now in the lab. Two comments and a question. The first comment is that being that I'm on the higher end of the BMI is good to know that there's some advantages to it. Not many, but at least kidney should not be one of my problems. The other one is that the clinical data that is in TCGA, the forms are extensive. Right. And they have been put together by mostly clinicians that were part of the disease working group. So if there's any factor that is not there, we can add them. The question is why don't we have more follow up is because the entities that provided the samples have not provided additional follow up. TCGA had payment for a single follow up, but we never closed the door. So people can continue to put the follow up in. And so my question in specific is, is your group willing to put all that data that you have accrued into the TCGA clinical site so everybody can use it? Absolutely. I mean, we were just finishing curating the follow up data, the treatment related data. And then, yeah, we plan on uploading all of that data. And specifically the cancer specific survival data I think is very relevant because, like I said, at least in kidney, 30% of the patients were dying from other causes. So that's very relevant when you're trying to come up with markers. Absolutely. And this is basically a message to everybody that has any clinical data. We have the door open. You can continue putting clinical data as long as you want until you're tired of typing. Yes, I think that's a great point. Thank you very much. So I'd like to follow up on the consortium we try to build that basically is really, really open access. And the other issue is we think this is a very good opportunity to do it because kidney cancer is all targeted therapies, mTOR inhibitors, or antigenic therapy. That's why we think we can actually obtain some insight from it instead of just the most of the cancer treatments chemotherapy. So it'll be more messy that way. That's why we hope that we can accomplish something. Thank you. Okay, the next speaker is David Wheeler from Baylor College of Medicine. He'll talk about the multi-center mutation calling in TCGA.