 So, it's been a pleasure to work with Steven, and we divided this up where he would be the rah-rah pep talk for a go team, and I would give an example from some of our own work that tells us by example what we could expect genomics can offer in the future to making cancer go away, which is, of course, our ultimate mission. And we've seen this, I borrowed this from Kenna, the pipeline of TCGA, and the part that I'm going to emphasize is some new boxes that we'd like to put in, that is, using clinical data in a very direct way, both the phenotype, the responses, survival, and putting that as part of the integrated computational analysis with the goal of changing clinical care, developing predictive markers of response to a drug, prognostic markers we already heard, is coming out of the TCGA, and then identifying new therapeutic targets. And of course, TCGA has had the requirement to get some clinical data, but this is more to actually putting the type of machine that is TCGA integral into a clinical trial, and I think that's the lesson that I'm going to drive home. Very quickly, I work for many years on diffuse large B cell lymphoma, it's the most common type of non-hatch lymphoma, and we can cure about half these folks, but we were unfortunately left with this current situation of finding new therapies for the other half, and I was interested in figuring out who was being cured by our current therapy, and that led to a subdivision based on gene expression profiling into three large groups, the ABC group activated B cell like, which we'll talk about a lot today, about 40% of cases, the GCB or germinal center, about 43% in a minor subgroup primary medistinal lymphoma, and it certainly is the ABC tumors that have the worst survival, or predict the worst survival, so we're maybe curing 35 to 40% of these folks, but we're doing much better, not good enough in GCB lymphoma at 75%. Now, early on, just looking at profiles, we saw a signature of NFKB activity, and this signaling pathway turned out to be essential for the survival of these lymphoma cells and culture, but we didn't really know why, and what we did was turn to functional genomics, RNAI screens, and developed over the several years the idea that it's an entire pathway leading from the B cell receptor on the cell surface to NFKB with the components shown here, and the first hit in our screen was a complex of signaling complex involving card 11 malt 1 BCL10, maybe not so familiar to this audience, but really well known to immunologists, and then by resequencing in a candidate way, we found that 10% of the ABC tumors had mutations that explained this requirement for card 11, that is these mutants were constitutively active when you put them into a heterologous cell, they spontaneously turned on NFKB, so that seemed to explain 10% of the problem, the other 90% cases had wild type card 11, but we had examples in the lab where they still relied on upstream signals from the B cell receptor, in particular if we knocked down a kinase BTK, they really didn't like it, but we sequenced all the kinases in the pathway and there were no mutations, and ultimately we found that it was the B cell receptor itself that when you disturbed any component of that receptor, the cells died, and you could actually see this, seeing as believing, so here is an image of the B cell receptor in the membrane of the ABC lymphomas, and these bright red dots are what you see when the B cell receptor is actively signaling in microclusters, and you don't see that in other lymphoma types. Now, given this clue, we directly sequenced the various subunits of the B cell receptor, and found in 21% of cases mutations that helped us understand this dependence on the B cell receptor, they occurred in two of the signaling subunits, CD79B and A, and they affected tyrosine residues that were very important for that signaling. Now, these were not as the same as the card 11, I would call them backseat drivers, if you put them in a heterologous cell, nothing happens, but within the context of an ABC tumor cell, they turn up the volume on B cell receptor signaling. Nonetheless, genetically this told us that this pathway was important. Now, more recently, it got complicated in an interesting way, where we came up yet another pathway that separately leads to NFKB activity, and that is driven by a signaling adapter called MiD-88, and I won't go into great detail about that, but we did find in 39% of this tumor type, activating mutations within one domain of MiD-88 that spontaneously, again, in a heterologous cell, will turn on NFKB. If you look at this wiring diagram, this makes some conclusions about where you might want to intervene therapeutically. Obviously, something the B cell receptor pathway might be helpful, but would the MiD-88 pathway somehow be redundant in a parallel fashion, so it would compensate if I inhibited just the B cell receptor pathway? Here's a clue from cancer genetics, and it's a point that I want to drive home also at the end, that we can get information about pathways, biological pathways, simply by looking at the raw genetics and the co-occurrence or co-exclusion of mutations. Here we found that, as I mentioned about, oh, I didn't mention this, there's one point mutation in MiD-88 that is the most potent one. It's called L265P, and it's found in 29% of cases. You ask, do mutant tumors with that L265P, MiD-88 mutation, do they overlap just by random with tumors that have CD79B or A? In fact, we find not, that there's a definite statistically important enrichment of those two, suggesting maybe they aren't redundant leading to the same goal, but could be cooperating. That led to a clinical trial, these kinds of findings, with a drug targeting the Brutus tyrosine kinase. It's a nice drug because it covalently modifies the protein, completely inactivates it for the duration of that protein's lifetime. We had found that it was very potent in killing our ABC cell lines in vitro, but did not kill other lymphoma cell lines, so it had pretty good specificity. At the NCI, with my colleague Wyndham Wilson, we did a short trial where we enrolled just patients with this ABC type of lymphoma and looked at 10 patients and treated them with this drug you take once a day by mouth, and has apparently no side effects that are significant for most patients. I'll give you two vignettes to show the power of this drug and something about the genetics. This woman had ABC tumor, she had inner tumor, the CD79B mutations, MiD-88 was wild type, and she had failed, this is all in relapsed refractory settings, she had failed multiple therapies, and went into a complete remission by week eight, the arrows point to some PET positive tumors that went away at week eight, and she is our star, she's come back to see us now, 2.2 years later, has no sign of disease, is taking this drug once a day by mouth, no side effects, and she's a happy camper. So we show, I'll show in second slide here, 59-year-old woman, wild type for CD79B, no discernible genetic mutation turning on the B cell receptor pathway, she had had a very primary refractory disease, very difficult to treat, her LDH in the blood was rapidly rising when we saw her, she had massive abdominal disease, and went into a very good but partial remission that she stayed in for only a month and a half, but during that month and a half she felt great. So this is the story of targeted therapy as we know, so I'll now give you a preliminary analysis of a completed clinical trial where we have 70 patients treated here, we took all comers, all DLBCL types, and use profile to figure out whether they're ABC or GCB, 70 patients, and our hypothesis was ABC patients should respond better, and that's what we saw, so 41% of the tumors of the ABC type responded only 5% of the GCB, so this tells you that molecular profiling works in the context of a clinical trial, but, and this just shows you more profound loss of tumor volume, and actually extensive survival, our patients are going out, this is still, as you can see, needing more follow-up, maybe heading past a year here for some of the patients. So for this audience, in particular, beyond the profiling, can genetic lesions identify the responders or not? Well, sort of. So here we have the CD79B mutation, small numbers, 5 out of 7 of these folks responded, somewhat of an enrichment over the overall response, that's fine, but look at this very healthy response rate in tumors with wild type CD79B, so this tells you right away that it's not going to be as easy as the stories with the BRAF and Vemeraffinib, it is not always mutation equals drug response, no mutation, no drug response. This is a gray area, and I believe a lot of what we're going to see. So then what about those Mighty 88 mutant cases? Well, if you had both a Mighty 88 L265P and a CD79B, you had a 4 out of 5 chance of responding in this analysis. A finding was that if you only had the Mighty 88 L265P, only 4 patients, but none of them responded, and then as predicted, if you had a card 11 mutation that is downstream in the pathway and you have a drug that's upstream, no response out of 4 patients. Now, we started to do an unbiased analysis of exome sequencing within this, and we had already noted that there was one very common lesion, which is deletion of the Inc4A-R flocus, specifically in this lymphomotype, and in red you can see that those, even within this already bad group, are the bad tumors. It's a finding. 5 out of 8 patients with that had responses, 0 out of 7 without that had a response. Small numbers, and we had to, so where are we going with all this? So a couple conclusions of how we can extend TCGA into this arena. First of all, I think I've learned that there's value in finding co-occurrence and co-exclusion of mutations, and I actually, if you sit down and ask your statistician to do the math, which I did yesterday, it turns out that if you want to drive this kind of analysis into the 5% and 2% range, you need 10,000 tumors. Actually, sometimes you can't even get there with 10,000 tumors, but if you want to see that 5% overlaps with 5% in a significant way, that's the number, and I think that would be a way cool project, but we got to come up with, Ken is trying to get 500, so we got to get 10,000, but I think there's real value in this genetic analysis. Second point, take a pathway-centric view of the genetic lesions. Look at both gene expression signatures, so don't forget about the phenotype, but also then use the genotype as well, and we may have to go to pathway modifications. And finally, we have to drive this to actual clinical utility. We have to make the predictive tests that are based on our genomic methods, make them available, and something that I think Barbara was the first to put forward, the idea of a cancer genome commons where people would freely donate their cancer genome into a large database along with their response to whatever treatment they had, and then any new patient might be able to sort of dial in their genome and see where they fit. So this is where we'd like to go. And it takes a team, it's a different team than TCGA, but it takes a team of researchers, both clinical and investigation, to pull this off. So I believe that the kind of integrated analysis that we're going to hear about in the next two days is exactly what has to be brought into, right into NCI clinical trials, and that's one of the things we're going to be discussing at NCI Central. So thanks a lot.