 Well, thank you and good morning. It's a pleasure to be here. I spent a lot of time with Dr. Passamani as suburban was transitioning, and actually it's also a pleasure to be part of the Human Genome Institute series on genetics and genomics. So I thought I would try, as the kickoff lecture, to just sort of paint an overview of why there's so much excitement, why there's so much interest, and just where we think new technologies in genomics, looking for my, sorry, will lead us. So during this talk, I'll first spend a little bit of time defining some terms, because all of us come from different backgrounds, and I just want to make sure the jargon is at least somewhat understood. Then talk about our vision for how genetic characterization of tumors will change treatment paradigms for cancer in the future, and with a description of some ongoing clinical trials. But before that, actually talk a little bit about what are the pitfalls and the difficulties. We can paint a very rosy picture, but we're learning there's a lot of more hurdles to be overcome, a lot more hurdles to be overcome. So let me first define a few terms, and then I'll show you some examples of what surveys of these look like. Single nucleotide polymorphisms or SNPs, I'm sure you've come across this in your reading or even in the newspapers now, and what we mean by that is a variation in a single base in the DNA, ACGT, most common variants in the genome, they're really over probably many more, but last I looked there were over 50 million SNPs identified. And SNP arrays are used to interrogate the entire genome, and usually we're looking at DNA from the germ line. So remember, there's germ line and somatic mutations. The germ line is something that's how the DNA looks in every cell in a person's body. A somatic mutation is when we're talking about a mutation that specifically happens in a tumor and you don't see in all the normal cells. And I'll show you, and these SNP arrays are often used in another term I wanted to find which is genome-wide association studies or GWAS. So they use SNP arrays usually to compare populations, and it's usually a disease state or not a disease state. Sometimes we use a GWAS to look at how certain drugs are handled, so there can be polymorphisms that affect metabolism of drugs, and that goes across all of medicine. And they're often used to determine risk or susceptibility to some state, mishandling of a drug, risk for cancer, et cetera. RNA expression profiles were one of the first sort of globally used genetic approaches because they could be done quite rapidly, and they're used to determine global messenger RNA expression, so that's the part that turns into protein, that's the code for protein, in a sample. And typically they've used hybridization of messenger RNA to a chip, and I'll show you an example of that. And that's led to the sort of oncotype and these kinds of analyses of tumors that are already in use today. Perhaps a slightly newer application of chip arrays are methylation arrays. So, so far we've talked about looking at the DNA or looking at the RNA, at the changes in the base pair and the coding sequences. Methylation arrays determine global methylation of the genome. So that's an epigenetic change in that it doesn't change the base pairs. What it does typically is that we have methyl groups inserted of what are referred to as CPG, so cytosine guanine, they're the CPG islands, and that methylation alters DNA. Typically it silences, but not always, but they are, so they're changes that we refer to as epigenetic changes, and it alters the transcription typically, and again it's hybridization of the DNA to a chip. And last but not least, which is really eventually going to replace all of these chip technologies, is the development of massively parallel sequencing, which allows for rapid sequencing of either the entire genome, which you sometimes see referred to as WES, or whole exome sequencing, or whole genome, which is of course, remember, most of the DNA doesn't encode protein. So when we do whole exome sequencing, we're looking for typically genes that encode proteins, but we can also do whole genome sequencing, or WGS, and of course we used to think of this as garbage, but we now understand this DNA has an awful lot to do with how genes function. We're still trying and figuring out how. And then there's also, you can take cDNA, which is a copy of RNA, and do what we call RNA-seq, but they all use the same approach, and I'll go through that technology briefly. So this is the way a GWAS would be reported, and this was actually, a data reported out a few years ago, when a GWAS study was used looking at SNP arrays, and this area is all around, this is around eight chromosome eight arm Q, so the long arm of chromosome eight in a region called Q24. So this is a distance across this particular region in centimorgans. And what you see are these outlying areas in red, which through these large population studies, identified a region, several regions, in eight Q24. So I put here, this is MIC, so MIC is an oncogene. So these regions were distal to MIC, but these regions were found to be distinct SNPs, so certain SNPs would exist in patients that had increased risks for either prostate and colon, breast, or prostate only. Two regions of prostate only, a region of breast, and a region of prostate and colon. And these areas were not encoding DNA sequences. So by interrogating the whole genome, we find these alleles, these polymorphisms that seem to be more frequent, statistically tracking with people that have increased risks of developing a particular cancer. Question is, how do we deal with that information? And the next slide shows the difficulty. So we've all heard of BRCA1 and BRCA2. So this shows the risk, so the population, the relative risk. So if you have BRCA, there's a huge effect, right? So you have a really high risk of developing breast cancer. But the frequency of that, this is the frequency in the population is extremely low, so the very potent polymorphisms or mutations in the germ line that increase the risk of cancer are very strong, but they're extremely rare in the population. What we more likely and typically see with these types of studies are alleles that are much more frequent in the population. So for example, the 8-2-4 allele, 30% of the population. But the risk is just over one. So the frequent alleles that we think predisposed to cancer have a intensity that's quite low. And what we're beginning to think, it's probably an accumulation of multiple of these alleles that ultimately we need to understand that will lead to better predictions of what risk factors are. But you can see mathematically, this is pretty complicated. So just showing you, these are very uncommon and rare, but they're clearly actionable today. These are much more common, but we don't know what the action should be. So this is an example of RNA expression profiling. And this is data from Lu Stout's laboratory over at the Center for Cancer Research. And he was interested in diffuse large B cell lymphoma. And you can see all of these tumors basically histologically look the same. And when he looked at RNA expression using these expression arrays, he was able to divide these into three distinct categories. And each of these red dots represents a particular gene that's highly expressed. And you can see this group of genes is highly expressed in a subgroup, about a third of these tumors. This group of genes is highly expressed in another third. And this group of genes is uniquely highly expressed in a much smaller subgroup. And Dr. Stout was then able to divide these into what he calls activated B cell and germinal center B cell. Partly because of the genes that are turned on. These genes more look like an activated B cell expression, what the RNAs look like. And this looks more like a germinal center B cell. Why is that important? What we're beginning to understand, and this is now just at the expression. This has nothing to do with the sequence at this point. Is that the germinal center B cells using current therapy have a survival of about 75%. Whereas the activated B cells have a survival of 40%. Histologically identical. So now we can begin to take action because this is unacceptable. And what Lou has gone on to do with a series of beautiful papers now using some sequencing, which I'll get to, is he's found specific mutations in pathways that have to do with NF-Kappa beta signaling. And there are some treatments, BTK, Bruton's tyrosine kinase inhibitors. And those are actually in clinical trial now. Specifically to get on that trial, you have to have an expression profile that looks like this. And this now can, once you have those genes, you can target, instead of looking at all the array, you can look at 20 genes, say, if these are on, I know you're an activated B cell type, and you can go on this clinical trial. So in summary then, with expression profiling, sort of two images of breast cancer. So this is what breast cancer looks like under a microscope, and this is what it might look like trying to look at all the expressed genes. And that led to what many of you are well aware of. We now know that breast cancer can be subdivided into luminal A, luminal B, HER2 positive, ER negative, and basal-like, or triple negative. And these are just four genes that you can see, can really separate, just looking for the expression. An oncotype, I believe, I should know this, uses 14 or so genes that can now allow us using otherwise similar tumors to now separate these. And we know these have vastly different outcomes. So I apologize, this is a little bit blurry. So this is data from our group looking at hypermethylation. So we were looking at gist tumors, and there were gist tumors that had mutations in C-kit, and there were what we called wild type gist tumors. And what we found is that the wild type gist tumors turned out to have a mutation in another gene in the Krebs cycle for those of you who remember your freshman biochemistry in medical school, the succinate dehydrogenase gene. So these were the normal gist that respond to a matinib. And you can see, hopefully a little bit, this is looking at a CGH, comparative genomic hybridization. This just shows that the genome has a lot of losses and gains. Unstable genome that we typically see in cancer, and it expresses SDH. This is a tumor that's SDH mutant, no SDH, and the genome looks like a normal cell. There's really no gains or losses. It really looks at the genome level to be normal. But when we looked at the epigenome, when we looked at methylation patterns, all of these tumors that look so quiet genetically had completely hypermethylated across their entire genome. We don't understand what this means, but these are the mutated tumors, and their genome looks pretty normal methylation-wise. These are the tumors that we can't find many mutations, but the methylation pattern is completely abnormal compared to normal tissues. So this is another layer of information. What about massively parallel sequencing or sometimes referred to as next-generation sequencing or NGS? So basically you can take DNA or RNA or CDNA, you fragment it. You collect those fragments, you put little pieces of adapters on them. And you do, it's called massively parallel because you're sequencing little bits of DNA that have been broken up and fragmented in these machines. And then you have to align each of these small reads on the genome, using a reference genome. But it allows you to sequence the entire exome if you want to just capture exomes or the whole genome. And these are the kinds of information you can get, and I just took this from a review a few years ago now by Matt Meyerson at the Broad Institute. And you can find point mutations. Again, you're always aligning to a reference sequence, so you can see point mutations. You can find insertions and deletions, which you'll see referred to as indels. So little pieces of DNA get put in, little pieces of DNA get lost. You can see whole deletions. So here you might miss a whole piece of the normal reference sequence. You can have that be homozygous or heterozygous. So you might find it in both alleles or just one allele. You get copy number alterations where you might have an extra copy of a piece of the DNA. And you can also, using the RNA seek, you can find translocations. So you find a piece of expression from chromosome one and suddenly it's hooked to a piece of DNA that lines up to a totally different chromosome. So all of that can be garnered from this massively parallel sequence. So here's the technology. So there's an astonishing array of machines now from various companies and various sizes. This eontorrent you can see is a lot, a lot of people are using this. You can put this on your desktop and actually sequence targeted areas. So to confirm, we use this often to confirm. Illumina is one of the big players in the market and they keep coming up with new generations. The current generation is called the high seek, but they also have something called a my seek. That's more like a personalized one to use in your laboratory. But this technology is changing at an astonishing rate. The point about this is only that they are extremely rapid. The standard operating procedures are getting very clearly defined and we can generate enormous amounts of sequence data very, very quickly. But just to give you an idea, for clinical practice currently, if we wanna do a whole exome sequence, because you have to analyze the data. So generating the sequence you can do in a couple days or a day now. Still have to analyze it and that the computing power is still trying to catch up. But it will take us two months or so from the time a patient walks in the door. So we're not yet able to take whole genome sequence and apply it within a couple days of a patient walking in the door. We can take targeted sequencing, 15 genes, 20 genes, but not the whole exome or whole genome yet. So I wanna go through now a couple of vignettes that I think will give you an idea of how this is impacting clinical oncology. So this is some data I took from Javid Khan, also in the pediatric oncology branch, a close colleague. And he was looking at sequencing the tumor that's near and dear to my heart, rhabdomyo sarcoma, which is a pediatric sarcoma of the skeletal muscle origin. And he did whole genome sequencing. So this is clearly an experimental approach. This is not involved in the clinic yet. From 46 rhabdomyo sarcoma is divided into about half or one subtype called alveolar and half or another subtype called embryo. He also did SNP arrays, which I just described to you. And then using this high throughput, he then validated all of the alterations found using a different technology, in this case either solid or the alumina technology, in a much larger number. So you can see he took 133 rhabdomyo sarcomas to see how this discovery set held up when he used a larger. So the first thing to point out is in the first 46 samples, two of those patients didn't have rhabdomyo sarcoma. And we figured that out by the genetic analysis. So one of these actually had a fusion of Alk and NPM1, which we know is probably anoplastic large cell lymphoma. Alk stands for anoplastic lymphoma kinase. So that was just somebody it's hard to tell sometimes. And there was another sample that we found another translocation, RET to NCO4, probably a parathyroid person. So some of the times we actually get a little bit more specific diagnosis. It was amazing to us though that that was only two out of 46 samples. So the pathologists do pretty well. So this is what we see. This is a circus plot. I don't expect you to understand or this is not the way data would be reported out in practice. But it gives you a sense of what you can actually get off of this analysis. So these are all the chromosomes, one through 22 and then the X and the Y. You can see this patient doesn't have a Y, so it's a female patient. And you can get, so this is the chromosome position around the outside circle. And then you can get copy number alterations. You can look at losses and gains. You can find translocations. So this is when you see a line drawn from here to here, that means this chromosome got hooked to this chromosome, in this case two and 13, and you can find copy number alterations and single nucleotide variants. So here is the translocation. We know this is common in alveolar, rhabdomyosarcoma, so we could find that easily. And the other thing was that there was this loss of heterozygosity. So here's looking at copy number and you see suddenly this is lost, it's a copy. This is loss of 11p15, so only one allele here. But the genome in general looks pretty quiet. You don't see a lot of wild changes here. So that's one type of rhabdomyosarcoma. Here's an embryo, and in this case what we see is much more changes. So here are all these mutations on the outside, I just labeled them here. So there are about seven or eight mutations. No translocation, and a few more copy number alterations. You see areas here in addition to 11p15, which is pretty typical across all rhabdomyosarcoma. So what did we find? We found the typical translocations. We also found a novel translocation as well as this Pax to a different fusion partner. So most of these packs, these two had been identified. We found a novel. We found three rhabda, alveolar rhabdos that should have a fusion that we couldn't find. We found something quite interesting. Here's a tumor where this chromosome two, rather than you seeing a nice line, there's this complete alteration of chromosome two. And what that looks like is this, so this was one of the ones we couldn't find the 213. And what we found is massive rearrangement of chromosome two. It just kept breaking and reconnecting and breaking and reconnecting. And so, and that's depicted here. So you see multiple, multiple changes. So we don't understand how that does the same thing as a fusion to a forkhead gene or fox01 gene, but these are some of the things we're learning. We did RNA-seq, as I mentioned, and in this novel fusion, in fact, it was expressed. So we could find the packs and then if you go along, it now starts reading sequence to this other gene on exon11. So question is, how does all this change what we do in the clinic? And so I'm going to now try to go to some common tumors. I do pediatric cancer, but most of you won't see pediatric cancers. But here's a slide, actually, I think I borrowed this from Charles Sawyer's. And this is something as common as lung cancer. So here's how we used to think of lung cancer. There was adenocarcinoma, squamous, and large cell. And in 1987, we knew a small portion, maybe a third, maybe between a quarter and a third, had KRAS mutations. In 2004, we found EGFR amplifications. In five years later, we now found these translocations, these out translocations, HER2 overexpressions, some BRAF mutations, some MET alterations, some AKT, some PIC3CA or PI3 kinase mutations. So now we filled up as many KRAS mutations with these other, but look what's happening. Each of these samples, now we're taking a common tumor. And we're making them rarer and rarer and rarer. So in a global sense, I sort of view this as really the whole, all of medicine, but in particular in oncology, the paradigm is shifting. We're going from descriptive medicine by understanding what something looks like under a microscope to understanding the mechanisms. No, it's not good enough to say this is a non-small cell lung cancer. Is it Alkfusion? Is it PI3 kinase mutated? And that helps go from empiric diagnosis to mechanism-based diagnosis. I'll come back to this in a little bit. And it's not even clear, ultimately, that everything will get grouped by organ site. It may be that we'll group things by disease drivers, although that's still unclear and I'll illustrate that. But we will certainly go from uniform treatment. This has already happened in breast cancer. It's already happened in lung cancer. The problem is what do we do with rare tumors like pediatric tumors where we're now subtyping them? It gets pretty difficult to do a clinical trial. We hope that by understanding the earliest changes in a tumor, we will have earlier biomarkers. And we can go from retrospectively diagnosing a disease to once we treat and we think we cure someone to be able to intervene before we ever see a change on an x-ray, but maybe when we start to see that. And this is, see that alteration. And this is, in fact, what's happened in CML. And hopefully, we can go from acute care to early detection and early intervention. So this is just a graphic depiction. And we like to call it, a lot of people call it individualized therapy. We like to call it precision therapy because most of us as physicians, I think, have been giving individualized treatment to our patients, our entire careers, at least I'd like to think I have. But we've not had precision tools. And so the idea is if you treat everyone the same and you see a responder, the question is, can we figure out what the lesion is and what we need to target and not treat the three out of four patients that aren't going to respond? And in addition, as I briefly mentioned, these pharmacodynamic measurements that may be identified through SNP arrays may tell us certain patients that would handle a drug differently so we can select drugs not just based on the tumor but also on how a patient will handle it, the molecular diagnostic approach. And again, so you can just see that breast cancer now will have two red types of tumor, one yellow type and three gray types. So that's really the theme. So now let me say why this has had such an early impact in cancer. Well, for many years I think it's been fairly well established that cancer is in fact a disease of the genome. So we all thought, okay, if we precisely define the cancer genome, we'll understand and cure cancer. We got to be cautious about this. And clearly this is a direction we need to go. It's clearly going to change the paradigm. But as usual in medicine and physiology and biology it's not so simple. So now I want to do a few more definitions because you'll hear about founder mutations. And what we mean by founder mutations is usually it's the first genetic alteration we can detect and it usually then is established and you'll see that in all tumors that are biopsied. And there are often lesions that lead to genetic or genomic instability like P53 or RB. These are, you know, P53 is called the guardian of the genome. So one of the problems is many of these founder mutations make the genome unstable. And they're often not fully transforming. Now we talk about driver mutations and what we mean by that is that those are required for the expression of a fully transformed phenotype and the driver mutations are those that we think we should be able to target and then successfully alter the disease biology. Then there are what we call passenger mutations and I refer to them as collateral damage. Their mutations that you find that just get accumulated because the genome is unstable but they're not necessarily, if you find that mutation and you block it, it may not have any function at all on the growth or survival of those tumors and so it may be noise. And so since most cancers are rapidly evolving because they have an unstable genome, it's really a major task to sort out drivers from passengers and I'll give you a few little graphic illustrations of that. So this is a slide that borrowed from Gotti Getz at the Broad Institute. And so he just sort of plotted here the frequency of mutation per megabase here across a number of tumors. So on the right is melanoma and on the left are pediatric tumors. And what you see is different tumors have higher or lower frequency of mutation rates. And it's interesting the highest mutation rates are in diseases that we know the environment damages DNA. UV radiation we know causes DNA. And you can see, I don't need to go into this but in the bottom it types, the colors represent the kind of DNA changes you get. So the yellow means you have a C to T transversion. And so there are different types of changes that occur but much higher in lung cancer. Again, we think smoking has a major impact. We know smoke causes DNA damage. And as you get across, here's multiple myeloma of varian cancer. So here's rabdomyosarcoma. So the genome there is pretty quiet but different types of tumors have different amounts. Some of these are enormously high. And this leads to the problem of sorting out what are the important mutations and what are just mutations that are not selected for that the genome is just unstable. So here's another way of looking at that. This, I think, I don't even remember what the tumor type was here but this was an interesting paper published from the Sanger Institute where they actually were able to look at various points along a tumor. And what's shown here is that there was an RB and a P53 mutation that occurred early in time. And by looking at tumors across time, they can sort of set up, here's what happened first, here's what happened second. And what you see are these branching chains. Each of these represent what we might call a private mutation that may just be occurring in this particular area of a tumor but not in this area, at the same point in time. You'll find two private mutations, what we call private mutations because they're only in one area of the tumor. So this is another problem. Even in one tumor, we find heterogeneity. And I'll show you some illustrations of that. Another point, another problem is, and we've learned this in spades, is that even when we find the driver mutations, they're usually in kinases, at least so far because we can target those. They're components of a highly integrated wiring that's not one way. And they're important for normal cell function and so they're highly regulated. So not surprisingly, when we perturb it, often the cancer cell figures out a way to get around it. And so this was an illustration of that. This is from a nice review that I would recommend talking about resistance in JCO, I guess two years ago now, by Levi-Garoway's group. So there are, here's an oncogene. You inhibit that oncogene. And often, so in CML or in GIST, what we find is you select for mutations in the same gene, so in the BCR-able gene, in the case of CML, that now doesn't respond to that kinase, but there's still mutations in the same gene. And what we think is happening is what we call clonal evolution. I'll come back to that in a few minutes. But we're really, those clones exist, they're much, they're low frequency. And by now killing the cells that are susceptible, we allow these clones that have already mutated, developed a point mutation that are now resistant to that particular kinase. But another defined mechanism is what we call a bypass, in which you block that particular pathway, and now the tumor just activates another pathway downstream of that. So it's not even a mutation in the same gene, just alter some other pathway. And that's been demonstrated now in Vemi-Raphanem, and I think that paper was actually defining that. Because for the first time, so these are melanomas, about half have mutations in BRAF, and now we're finding these mutations much rarer in other tumors, but it's a mutation from V to E. It's called a V600E mutation. And what we learned is that doses of this BRAF inhibitor that inhibit 90% of this kinase activity, most patients, the overwhelming majority of patients respond quite rapidly with tumor shrinkage. Unfortunately, the overwhelming majority of patients recur less than 12 months later. And so what we've learned is that they're not mutating their BRAF like we saw with CML. They're activating other downstream pathways. And another interesting thing, remember that part I talked about, we might treat, instead of by organ site, we might treat by BRAF mutant tumors or HER2-amplified tumors. Well, here's a word of caution. This was a phenomenal paper published a year ago by Robards Group and Bernard's Group in the Netherlands where we're beginning to find BRAF mutations scattered throughout, and so they found a fair number of BRAF mutations in colon cancer. And the idea was they should respond to Vemiraphinin, but they didn't respond. And it turns out when you treated these patients with the same drug, Vemiraphinin, they activated EGFR, so epidermal growth factor receptor. And the mechanism appeared to be that BRAF leads to inhibition of this other pathway called mechanurk, and that changed the phosphatase activity which then shut off EGFR. So now when you perturb this, now you up-regulated EGFR, and that then was driving the tumor. So the reason I spent a little time to talk about that is that it's not gonna be so simple as treating every defined mutation, and if it occurs in colon cancer, it'll be the same as if it occurs in a melanoma. So context is gonna matter because the wiring is gonna be different depending on what part of the, you know, what's the histology and what organ you're looking at. So life is getting more and more complicated. Now I wanna come back to this idea of clonal evolution because this is a whole nother, so we talked about resistance, so let's talk about clonal evolution and how that's also a difficulty. And this is, again, data I took from Jabid Khan. So he's also been sequencing neuroblastomas, very rare tumor, and children over one, it occurs in young children and it's got a horrible survival rate, less than 30%. So we were very motivated. We thought, let's understand the genome and let's find new ways to treat this tumor. So this is actually an older patient. A patient that had high-risk neuroblastoma, he had metastasis in the bone marrow, and at diagnosis his bone marrow was chock-full of neuroblastoma, so that was easy access to tumor, unfortunate for him. He had this big tumor in his liver and his primary tumor in his adrenal. He was treated with four cycles of induction chemotherapy. He got surgery, so we were able to look at the primary tumor, and the tumor unfortunately was viable at the margin. He got additional cycles of chemotherapy, did well for a few years, and he ultimately died, and we were able to go look at a second metastasis, a second site at the time of death when he died. And so we did whole genome sequencing of the liver metastasis. We did an RNA sequencing also of the first metastasis and also of the primary tumor and the metastasis. So the first thing is that in the liver metastasis, 44 mutations were found. So how are we gonna deal with 44? And look at all these translocations here. So here's all the mutations, there's lots of translocations, lots of copy number alterations. So this genome was quite abnormal. Actually something called chromothripsis had occurred, and that's due, it's almost like the chromosome blows up. And we see this now, this has been well described, and in this particular liver met, it had happened in chromosome four and chromosome 13. And the chromosome was just massively rearranged as if it just blew up and got hooked back together, obviously not in the right order. So then we re-sequenced using the eontorrent of the primary bone, so all the abnormalities and the bone marrow, the four areas of the primary tumor, and then the metastasis. And the first thing was we found differences, there were small variants in four areas of the same tumor, we could find some differences, and that's been described now in kidney cancer. So that's another theme. You get a big tumor, you look at the North, South and East and West poles, you'll find specific mutations that are what we call private to those parts of the tumor. The good news is there are common mutations, but if one of those new mutations can be a driver, then we have to treat all of those. And so here's 14 of those 44 mutations were found across, this is looking at the met one, the primary tumor and the bone marrow, or met two in the primary tumor. So the good news is a third of those were really common. So those are probably the ones we would go after, but the bad news is a lot of them were private mutations in each of those individual tumors. We found three genes that were high in all these tumors that might indicate these could be targets. But the bad news is 30 of the 44 mutations in the second metastasis that was so genetically unstable were only seen in that liver metastasis. So, and some of these could be drivers, we don't know. But this is the kind of heterogeneity that was a nightmare for an oncologist. So neuroblastoma, and remember I showed you neuroblastomas at the left end, the low mutation rate of cancer. So there's every reason to believe in other tumors it's gonna be more complicated. It's marked by aneuploidy and we mean by that a messed up genome in recurrent regions, but frequently the mutations are not recurring, they're private. It's possible that some of these mutations may drive tumor genesis, but it's also possible that each individual tumor has its own set of drivers and that's a nightmare. And ongoing efforts are now, we're trying to see what are the commonalities and if could we find two or three that if we could target all three would take care of all of this. And we just don't know the answers to that yet. Another problem is that sometimes we find drivers and remember I mentioned these epigenetics. So these are things that alter epigenetics, they change, they alter methylation of either the genome itself, the CPG islands. Sometimes we find things that alter the methylation marks of a histone which then determines whether a gene is gonna be activated. And all of these in red are mutations that are clearly found as drivers and here's the histologies they occur in. And we're finding these more and more frequently and we don't have targets for those. We don't really know how to target many of these mutations. So resistance is a problem, clonal evolution is a problem and finding mutations that are drivers that we don't have drugs for is another problem. So I wanna end with a study that's just about to open across the street. Partly it's an advertisement, we'd love to have patients referred for this study. So it's really now up to all of us to ask, we can find these alterations, we can find alterations that we have drugs for. What's the evidence that treating those is actually better than doing what we've done for many years which is take the combinations that we know have some activity and treating them. And of course we have to ask those questions ethically so we have to start with advanced tumor patients but the question is, can we objectively show in cancer? I mean we clearly know that when we find a driver mutation like a BRAF mutation, patients respond but they all recur. So can we really show that the use of this expensive analysis at some point during the lifetime of a patient's tumor that it's worth having that information and that we can act on. And so the objective of this study is to assess whether response rate and four months progression-free survival is improved following treatment with agents chosen based on the presence of specific mutations and only patients with predefined mutations will be eligible and I'll show you that in a minute. And the studies treatments will be standardized and chosen from a list of regimens in the protocol and I'll describe that in a minute as well. And arm A will then receive the treatment based on identifying a mutation and saying, yeah these are the agents that should target that and arm B is just we're gonna pick one of the other arms. So here's the patient population pretty standard so they have to have refractory solid tumors, normal organ function, a standard early phase study. And we're looking at mutations in DNA repair pathways and if we find those, depending on what the repair pathway is, there's a PARP inhibitor plus an alkylating agent or a WEE1 inhibitor plus a carboplatin. For those that have mutations in the PI3 kinase or loss in this pathway, PI3 kinase AKT, they will be treated with an mTOR inhibitor and for those that have mutations in RAS, we'll treat with a MEK inhibitor that's just downstream of RAS signal. So here's the study design. Everyone has to get a biopsy. The patients are sequenced. If no mutation is detected and obviously we're targeting, so we're not doing whole genome. We're targeting those pathways, DNA repair, AKT. And so we can do this quickly. If we don't find a mutation in one of those pathways, they're off study. If we find a mutation, they're then randomly assigned to arm A or B and the clinical team is blinded to where they're assigned and then they'll either get targeted therapy or they'll get the alternative for one of the other targets. And there's an allowed crossover because obviously, you know, if somebody gets randomized to a combination that doesn't target their pathway and they progress, we'd still like to give them the opportunity. So they'll be randomized two to one with more favoring the targeted therapy. Up to 30 patients will be treated. Two arms will be compared with respect to objective response and progression free survival and it's a randomized. So here are the pathways that we're looking at. So this is the RAS pathway. So we're looking for mutations here and they're either gains or losses but these patients will be treated with MEK inhibitors. Here's AKT, PI3 kinase P10. So there's either gain or loss and here they'll be treated with an mTOR inhibitor and here we're looking for DNA repair and depending on which gene is mutated, we'll either treat them with a W1 inhibitor plus carboplatin or a PARP inhibitor plus temozolamide. And I should have thought I had this but I didn't show this. So the patients, so let's say we find a patient with a RAS mutation. Two of those patients will get randomized to be treated on the MEK inhibitor and the other patients will either get, will get one of the other, the regimen determined to be targeting this or this so that everyone's getting one of three treatments but either it's specific or it's chosen because it's not specific. So I'm just gonna conclude with a couple of sweeping generalities. Certainly we believe that the ability to obtain full genomic data on a given tumor will allow us to make much more rational choices for therapy. We hope that functional genomics may provide help in choosing combination therapy and I didn't have time to go through functional genomics but it's a way of just when anticipating resistance I believe by when we have a targeted therapy you can look in vitro and see that other pathways get activated and by using specific inhibitors of pathways either genetically or with drugs we can predict what gets activated and then maybe be prepared as a patient develops resistance to add on additional therapy or even start with combination therapy. I believe that we've done most of our studies to date using single agent targeted therapy and as many of you know as I've been practicing oncology for 30 years now single agents don't cut it we never tried to cure anybody with single cytotoxic therapy. We know that we have to combine targeted therapies and in my mind we need to get much faster at getting to those combinations but combinations are not panacea either because we already have learned that some combinations just like with chemotherapy have too much toxicity so it may be a perfect combination except the patient doesn't tolerate so we're gonna have to learn we have to be quicker at getting to combinations and I would just end with saying I think in the end our hope is if we can't cure patients we can turn these into chronic diseases and that's not such a bad thing as long as we recognize the rapid development of resistance and clonal evolution it may be that we just continue if a patient responds and then develops a resistant clone maybe we don't stop the therapy they're on maybe we just keep adding things assuming that they tolerate it and so that's sort of the hope for the future I hope I've given you some sense of where genomics is today where we hope to take it in the future and also given you a realistic picture of we got a lot of work to do thank you very much for your attention I have a couple of questions about the tumor I have one of the promo triceps that you mentioned is it possible to have the function of chemotherapy? That's a good question I believe occasionally these have been found in de novo tumors so I don't think it's a function but it's possible that it's more common in the setting of chemotherapy I don't think we have enough data yet Other comments or questions and could you paraphrase these for us? So the question was for the last study I described are brain tumor patients being accepted I believe so so the PI of that study is Shivani Kumar it's K-U-M-M-A-R first name is Shivani you should be able to find her I should have put up a contact phone number in fact I can find you afterwards I have it on my iPhone so I can give you her contact she'd be delighted to take any calls for referrals study isn't open but hopefully it's going to be open very shortly Other comments or questions I have another which has to do with the pharmacogenomics it seems like that post the clonal stuff and the resistance really gives you a very complex problem to solve do you comment? That was the point I was trying to make I think we're going to have and of course all I talked about was the tumor there's been a lot of interest in interactions between I don't like to use the word host because I think it's a little bit of a messy term but the normal stroma surrounding the tumor and how that may be influencing certainly we know there are signals going from the normal surrounding stromal tissue to the transformed tumor tissue and there may be changes there as well certainly the pharmacogenomics I think we're just scratching the surface and that may alter our selection of certain drugs as we learn which drugs may be more toxic in a certain SNP background or genetic background than another so this is complicated that's why I think our initial blush of we're going to sequence the genome and find the cure for cancer is a little bit more optimistic than realistic but there's no question it's going to change our treatment