 Good morning, everyone. My name is Ali, and I'm a computational biologist from Dr. Levi Garroway's lab. Today I will speak about somatic alterations in clinically relevant cancer genes among 12 TCGA cancer types. This presentation is based on a project that was done in collaboration with Dr. Han Liang of MD Anderson and is currently in press. Okay. So I don't need to tell the people in this room how much TCGA has helped identify alterations that may contribute to many different tumor types. However, less attention was given to this data with a clinical lens. As these profiling technologies that enable TCGA move into the clinic, an understanding of the landscape of clinically actionable somatic alterations may inform patient care. So now the question is, what does the spectrum of alterations in clinically relevant genes look like across tumor types? Somatic mutations and indels were called from 3,276 TCGA patients, and we focused on clinically relevant cancer genes. So what are these genes? From our perspective, these are genes that when somatically altered can predict response or resistance to known therapies, or it might have diagnostic or prognostic utility. So we created the target database, which is a set of genes that we define using expert opinion and literature to be clinically relevant in cancer. This database is online as of about 10 days ago. And here's a link to it. At the time of the analysis, this list contained 121 genes. We know that this is already out of date, since it's very difficult for a single group sitting in one office to keep track of them. And we think we could do a much better job as a community. So we want to encourage you to contribute to this list if you have any genes that you would like to nominate. Now among these 12 tumor types, we identified over 600,000 mutations or short insertion deletions. And of these a little less than 500,000 were non-synonymous. Of these non-synonymous mutations, over 10,000 alterations occurred in clinically relevant cancer genes in about 90% of the patients. Now we do recognize that not all alterations are drivers or are known to be clinically relevant, but we do want to highlight them to promote further study of these alterations and to better place them in a clinical context. Here you can see the frequency of alterations across the 12 tumor types in the clinically relevant genes. I know it's impossible for you to read the names of the genes, but what I'm trying to demonstrate here from a 30,000 foot view is this long tail of alterations in clinically relevant genes. Now you might be wondering what the highest peak here is. It is TP53. Before you roll your eyes, I want to say that we do recognize that it remains unclear what the clinical relevance of TP53 is from a therapeutic standpoint. But it remains an entry criteria for many clinical trials. Therefore, we feel that it's worth including here. There has also been some very interesting clinical trials emerging with neuropneum. Recent evidence shows that patients with ERB2 mutations show response to neuropneum. I've also included a couple of other mutations, but rather than going through every single gene here, I would like to highlight a few key themes that emerge from this analysis. So theme one. Right now, many clinical centers are using hotspot profiling technologies for clinical use. These include technologies like mass spec that find individual predefined events in a given gene. Here we have taken a representative hotspot profiling approach called Oncomap that looks at over 400 alterations in 41 genes. And we ask by tumor type with this technology how many patients have at least one alteration in this panel. As you can see, this does cover a reasonable territory. Now if we perform whole-exam sequencing and take the same set of genes and ask how many of our patients have alterations within the same list of 41 genes, as you'd expect, we do find a lot more patients with alterations in this gene set. Expanding next-gen sequencing into the clinic will help us identify more clinical alterations than hotspot profiling. This can help us to take more clinical genotype-to-phenotype relationships. Now one might argue this objective might be achieved if you just looked at targeted panels. But you have to keep in mind that the list of clinically actionable genes continues to increase. And the cost of doing whole-exam sequencing is approaching that of doing targeted panels without the need to re-sequence as information about new genes emerge. Theme two, hotspot on other alterations in well-known clinically relevant genes occur rarely in unexpected tumor types. For example, as expected, we observed BRCA2 alterations in ovarian and breast cancer. However, when excluding patients with mutation rates of higher than 10 per megabase pair, we observed BRCA2 alterations in multiple tumor types. These alterations may have clinical relevance and are generally not thought of in these tumor types. As these mutations may predict sensitivity to PARP inhibitors or cis-platin-based chemotherapy, so they may have clinical relevance. For research purposes, this may not be a significant gene in this tumor type. But for the few patients with this alteration in a cohort of, let's say, 500 patients, this is clinically significant and needs to be identified. Theme three, genes that are rarely mutated in one tumor type occur frequently when considered across several tumor types. As an example, TSC1 alterations may predict sensitivity to mTOR inhibitors, at least in bladder cancer, as was demonstrated by David Solid at Memorial Sloan Kettering. However, having a clinical trial in only one tumor type is not feasible. But if you look at TSC1 mutation status across tumor types, you can imagine a clinical trial. These types of trials, the so-called basket trials, are emerging. And this is just one of the many examples where genes rarely mutated within one tumor type become frequently mutated across tumor types. In addition, we also recently learned that certain mTOR mutations may predict sensitivity to mTOR inhibitors, like Everolimus. As with TSC1, mTOR mutations occur rarely in any given tumor type. But in aggregate, they provide a rationale for a similar basket trial. Thus, if you consider low-frequency alterations across tumor types and within pathways, one can further imagine clinical trials that can be designed with this knowledge. So our ability to increase the link between alterations in cancer genes and clinical trials is improving. Here, we downloaded all of the clinical cancer trials from clinicaltrials.gov from 2005 to 2012 and asked whether they had the name of a clinically-relevant gene in their title. You can see that the number of trials that incorporate clinically-relevant genes continues to increase. Now that we've identified the landscape of alterations among a list of clinically-relevant cancer genes, we have the opportunity to design the next wave of rational clinical trials. So in summary, what we saw today was a long tale of alterations in clinically-relevant genes. We also saw that going from hotspot profiling to exome sequencing will yield a more complete and clinically-useful patient-tumor profile. We also showed that clinically-relevant alterations in well-known genes occur rarely in unexpected tumor types. And also, genes rarely mutated in any given tumor type are more regularly altered when considering aggregate studies. Broadly, we know this is just a start. To improve on these analyses, we are moving to integrate additional data types that may have clinical impact, such as copy number variations or targetable fusions, such as ROS1 fusions in lung cancer. We hope that this work will kickstart more omic studies with a detailed clinical focus. I would like to thank Dr. Garaway for his mentorship and guidance, Dr. Van Allen and Dr. Wagley, Gadigets and Mike Lawrence from the CGA group, Dr. Han Diyang and Yuan Yuan from MD Anderson and Lars Sononberg. Also, I would like to thank the TCGA organizers for giving me this opportunity to present. I have a few minutes if there are any questions. Hi. In theme two, where you showed mutations in BRCA gene appearing in other unexpected cancer types, were these mutations that were recurrent across cancer types or with a specific mutation specific to those unexpected cancer types? So, I'm talking about position-wise, are these hot spots or are they? To my note, no, they weren't. They occurred in different positions. Do you think that's important? Whether they occurred at the same position or not when you're looking at across cancer types? Well, it might be more interesting. What we want to say is that it's worth taking a look at. I don't know whether they're interesting, whether they're useful information, whether they're driver events, but it's worth taking a look at. Thank you. Thank you, Ali. The next presentation is by Leila Esper on inferring intra-tumor heterogeneity from whole genome, whole exome sequencing data.