 Good morning, thank you for allowing me to speak today, I just want to say that it's a real privilege to be part of the TCGA, this is a very impressive symposium with everyone in the room, I think congratulations go out to Kenna and Brad and Elaine and Linda for really putting on a really very impressive first symposium, and I've been very privileged to be part of TCGA for a number of years, and so, I'm getting called. So today I'm going to talk about validated targets associated with curatively treated advanced serious ovarian cancer, I just want to point out that Joyce Barlin in my lab did a lot of this work, and we used a lot of TCGA data as well. So when you think of cancer genomics, I think a lot of the talks we've been hearing about of course are cancer genomics, and I see cancer genomics as discovery and technology development, we've obviously heard amazing stories and amazing information in the past day and a half. We try to do applied cancer genomics. So as a surgeon scientist I like to look at a question that I see in the clinic and try to see how can we apply cancer genomics to answer that question. And in this way, TCGA is very well suited for applied cancer genomics due to the richly annotated and well designed collection of clinical data, and I think all of the applied cancer genomics will really improve in parallel with all the tools that are being developed, and as the access for the mere mortals becomes easier to get to. So this is ovarian cancer, and basically the way ovarian cancer is treated is that it starts with a, usually with a big operation, and the goal is to remove all the cancer that you can see, and then patients get chemotherapy, and then time goes on, and hopefully the traditional or historical five years goes on and patients are still alive and they're very happy and we sort of consider them cured, but certainly even if they're not cured they're living well and they're living long after this type of approach, but it doesn't always work that way. Sometimes you do an operation and you can't take out all the cancer and they get chemotherapy and they die very quickly, and those patients have platinum refractory disease. And more commonly, you know, this goes on and patients do well and the cancer comes back and you have recurring cycles of coming and going and then you die. And then sometimes you'll have a long period of time where you're doing well, you get another operation and maybe you get some more chemo, cancer comes out, cancer comes back, and you're still alive five years later and you may or may not be happy because you still have cancer, yet you're alive so you're doing pretty well. So you know, many platforms can be used to address these types of questions, and certainly gene expression has been used more commonly than others, and a number of studies have been done looking at the comparison basically of long term survivors and short term survivors. This is a paper from Andrew Birchuk from Duke who's in the audience here, and they picked short term survivors and long term survivors and did some gene expression profiling and then developed a predictive classifier and they can pretty well classify patients who live long or short, and they identified a gene called MAL, which is involved in myelin biogenesis as well as lipid membrane maintenance, and they feel that that may be related to platinum resistance and I've tested that in some subsequent studies. So that's one approach. This is the typical cancer patient here, this is the TCGA data, gene expression clustering, some predictive signatures were developed, they were validated in multiple other studies including the Birchuk study, one of our studies, the David Botel study, he's in the room as well. And you can divide patients into different groups and you can sometimes put your thumb in between the lines, but for many of the curves at the end, the patients end up in the same spot, which is not the spot where we'd like them to end up. And so every so often we see this, and we don't see this a lot, but every so often you see something like this where there's this type of patient who has all these treatments that I mentioned, and then here you notice the curve sort of flattens out, and this is nine years, and there's people on this curve, however we define this curve, who are living a really, really long time, and I'm not sure if they're cured, but they're certainly living better and longer than the people who are on these other curves, which is unfortunate what we see all too commonly. This is a study from my center, this is a study from Europe, and so there's always every so often there's these types of patients. And so we were wondering, and I'm sorry, and so we call these patients, for better or worse, we call these patients one and done patients because they get one set of treatment and they're done, which is in contrast to the traditional patient who has treatment, has relapsed, has more treatment, more relapsed, more treatment and dies. And so these patients got one initial upfront treatment of surgery and chemotherapy, and they're done with their treatment, and they're alive five years later without any recurrence and without any evidence of disease. And so this is a small subset of advanced civilian patients that are apparently cured, I don't know if they're really cured, but they're apparently cured after this treatment, and our question is how are they different from other long-term survivors who have recurred? We don't want to compare them to the short-term survivors who are platinum resistant and have a lot of other biologic mechanisms going on. We want to compare them to kind of a similar group of patients because we don't want to study platinum resistance. A lot of people are studying platinum resistance, it's very difficult. Maybe we tried to pick an easier question to study, but we wanted to sort of exclude that subpopulation of patients, and our hypothesis was that expression profiles would be different between these groups. And we started this about a year ago, and it turns out it wasn't a bad question because one of the 24 provocative questions released by the NCI a few months ago was, I can't read it from here, but why are some patients with disseminated cancer apparently cured with chemotherapy alone? Our question is similar, ovarian cancer is certainly disseminated, yes. This population is sort of cured by chemotherapy alone. They have surgery with chemotherapy. I'm not sure they're cured, but they're living a long time. So this seemed like a reasonable question to ask. And so we chose our patients, we chose our two groups, and then we looked for available data. And since there's very few patients in this long-term survival group, we had to go to multiple sources. And fortunately, we had our own study, we had done it at Memorial, we had access to the TCGA data, we could remove any overlapping patients there, and you don't end up with a lot of patients. So in the case group, really there's 14 patients that we had and only 16 patients in all of TCGA from the clinical data that's in the paper. And these formed our two groups, and they were done in slightly different time periods and batches. And so we didn't want to just combine all the data, we actually kept the data separate. And so our first analysis of the gene expression data, we did standard, first of all, unsupervised clustering in the TCGA data on the right. You can see there does appear to be one clade where most of the long-term recurrent patients reside, and not a lot of these one-and-done patients reside here. But it's not really so informative, the unsupervised clustering, you don't see any beautiful clusters. We did a supervised analysis here, and in every whatever threshold you choose from our filtered set of genes, you certainly find many more genes than you would expect by chance alone, and the heat maps suggest that to some extent. But both data sets suffer obviously from very small sample sizes, but they are done independently. So we wanted to overlap this data and see what was similar. However, we first did a pathway analysis with IPA, and in both data sets, as often happens in IPA for better or worse, NFCAPB transcription comes up, ERC signaling comes up, these were both overrepresented in these two sets of data. And so then we chose, first we chose technical validation because these were both done on arrays. We wanted to validate them with an orthogonal approach. So we used nanostring, which we found to work quite well. We actually picked 19 genes that overlapped between the two data sets, and then we selectively chose genes from the pathway analyses, and some of the genes are listed there on the right side. And many of these genes, these are the genes actually that validated. And then of course we have to do one more iteration here. And so we go back into now our archival collection because all the frozen material from our center from TCGA have been used in the initial analysis, but fortunately we can now access the archival material. We were able to identify 57 additional patients, 25 fell out into the one and done population, and 32 were in the long-term survivor population. And it turns out if you look at the genes that have survived this sort of third or fourth set of validation, you really end up with three genes down here. And it turns out these are the three genes that were found by overlapping the two initial data sets and all the genes that we tried to pull out of the pathway analysis, maybe because it's not such a great tool or we're not so good at using the tool, whichever it may be. But none of those pathway derived genes survived the second set of analysis. And you can see here that in the recurrent patients, there's a greater expression of this cytochrome P450 enzyme. And so we are now beginning to move on to doing some biology behind this gene. It's a cytochrome P450 enzyme. It's really not well studied. We do know it participates in various drug metabolism in the liver. It's also involved in cholesterol and other lipid biosynthesis and a number of genes that have been thought to be associated with drug resistance specifically in ovarian cancer are thought to be part of the lipid membrane and also lipid biosynthesis. We actually think that because we've sort of excluded the platinum resistant patients, this may be more of an effect that we see with taxanes. Taxanes are metabolized by the liver. You could make the hypothesis that the patients who are not able to metabolize the taxal as well get more of the drug effect. And therefore that's the reason why they're doing well and surviving longer. We're going on, like I said, just to start, just to start these sorts of biologic studies to really try to figure out what is unique about these patients and because they're living a long time, I guess as I just mentioned, we think this may be more of a taxane phenomenon than a platinum phenomenon. And so, oh, so then of course we go back to TCGA and see what else can we learn about this gene in general, not just from our focused selection of very well-defined groups of patients. From the C-bio cancer genomics portal, you can see that this gene is amplified in 4% of the cases associated with increased expression. But when you look at the methylation plot, there's an interesting subset right down here. Again, I would just guess about 5% of the cases that are methylated with decreased expression. And then if you look at the overall survival and there may be some over-accounting here, but again, you find the patients who are amplified, you do find a few of them who have this long-term survival feature that we're seeing here. And so, you know, this one particular gene seems to be overexpressed in this specific population. I just think this is a nice example of how we can take the TCGA data integrated with other ongoing questions that are being asked at laboratories outside of TCGA. I think I'm sure many people are doing this as well. And of course you have to validate this which we're doing with orthogonal methods since doing the biologic studies. But it's very nice to take TCGA data to either integrate it to increase power. Of course people are gonna use it to validate their initial findings and to create a new discovery. And it's just such a well-suited resource for applied cancer genomics. And so I thank TCGA, NCI, NHGRI, and everyone in the room for really creating the tools and creating the data and making the data available and making the whole project, you know, move forward. And of course the generous patients who are generous with their signatures and their tissues and their willingness to share their stories for our research. Thank you. Questions for Dr. Levine. Doug, great presentation. We've learned that BRCA1 and 2 carriers tend to do a lot better than the average patient and can be long-term survivors. So I wonder if you had the BRCA1 and 2 germline mutation status. We don't have that from our own population because that's, we can't get that information easily. So it would have it for the TCGA group. But that also didn't come up in our initial analysis. Not that you'd see it in an expression. Wonderful, love presentation. I didn't catch, did you state the stage of these patients? They're all advanced stage. They're all stage 3C and 4. They're all high-grade serous. It's the same, our cohort basically is the same as TCGA. And so it's that, whatever's in the nature paper. Okay, thank you very much.