 So I just want to thank the organizers for giving me an opportunity to speak today and hopefully my talk warms you guys up in this chilly auditorium. So my name is Ian Watson. I'm a postdoctoral fellow in Dr. Linda Chin's lab helping lead the melanoma TCGA analysis which is co-chaired by Dr. Linda Chin and Jeffrey Gershengwald and I just wanted to put up all the people responsible for the different platforms. This is quite a, like many TCGA projects, a very collaborative effort. In particular I want to point out the efforts from Lee Hwa-Zoo and Terence Wu who are data and analysis coordinators and also thank the fire hose team. Our point people have been Dan Dakara and Mike Noble for all the custom adjustments they've made to fire hose given that the melanoma TCGA project is different from every other TCGA project to date in that over 80 percent of our samples are from metastatic origins. They originate from non-glabrous skin primaries so that's the hair-bearing parts of your skin. We've excluded samples from the palms of the hands and soles of the feet as we know that they have different genetics from past studies. Another criteria was no prior systemic treatment. And the rationale for this is that with primaries, they're very small, they tend to be nevi so you don't get a lot of tissue in order to carry out all the different analysis for the TCGA project and most of those samples goes for diagnostics. Also if discovered early, melanoma is highly curable, however the survival rate decreases significantly for regional disease. And what we're studying here is the most common first site of metastasis in melanoma. So the majority of our samples are from the regional lymph nodes as well as regional skin and soft tissue. We also do have a good number of primary and distant metastatic samples which will allow for some unique TCGA analysis in the future, however I should note that there are only two cases with matched primary and metastatic samples. We stop the data freeze at batch 291. We have up to 330 samples for a given platform and the rate limiting platforms for the integrative analysis like many TCGA studies are the RPPA and low-pass hold genome sequencing. So we know that melanoma has the highest mutation rate of any cancer to date and UV plays a significant role as demonstrated this nice figure from Mike Lawrence and Gadgetz showing the comparison of the mutation rates across all cancers. We confirm this result in that we observe a mutation rate of 17 mutations per megabase which I think is the highest reported for any TCGA cancer to date. The majority of the mutations are C2T transitions at dipyramidines which is consistent with the role of UV induced DNA damage and misrepair. We are left with the problem of identifying driver mutations when we've identified over 200,000 single nucleotide variants in our up to 300 samples. So how are we going to address this? Well first we turn to mutate and we identified 42 significantly mutated genes with a Q value of less than 0.1. We also used INVEX which is a statistical tool which also takes into consideration intronic mutations to try to determine a gene specific background mutation rate and has been used in past studies at the Broad Institute in a collaboration with Linda Chin, Levi Garroway, a talented computational biologist, Iran Hodes and myself in a previous analysis of exome sequencing data for melanoma and was used in a really nice study by Matthew Meyerson in another high mutation rate cancer. So what do our significantly mutated genes look like? We identified 13 significantly mutated genes by INVEX and these 13 genes were also found on the mutate list. What gives us confidence in this list is we identify the known melanoma drivers such as BRAF, NRAS, CDCAN2AP53 and P10. We also identify significantly mutated genes that have been found in a few exome sequencing studies including PPP6C, ARID2, MAP2K1 and RAC1. We also identified for the first time to be significantly mutated melanoma associated genes NF1, IDH1 and RB1. So they've been implicated in melanoma but here we find the first time to be significantly mutated given our large sample set and we identify a novel gene DDX3X which is RNA helicase. So this is our landscape figure. You can see the mutation rate at the top, the samples are indicated in the columns. We see the mutation spectrum below and as you can see the majority of the samples do have a high fraction of CDT transition indicated of a roll of UV. One thing that's clear from the landscape is that we do see a high fraction of BRAF hotspot mutations which is known. Majority of them occur at the V600 residue. These are mutually exclusive with hotspot mutations in NRAS and the majority of them occur at the Q61 residue. So one of the major questions in the field is what's driving these BRAF NRAS wild type samples and for the first time we've sequenced enough samples to really get an idea of the landscape of mutation in this subset. One thing I want you to notice is that these BRAF NRAS wild type have a higher fraction of samples that lack the UV signature and in the samples that have the UV signature we see an accumulation of these NF1 loss of function mutations. So the BRAF hotspot and NRAS mutations are significantly anti-correlated but also the BRAF hotspot and NF1 mutations are significantly anti-correlated. So we only observe one co-occurring loss of function NF1 mutation with a hotspot BRAF mutation. So what do these NF1 mutations look like? Over 50% of the mutations are loss of function mutations either non-sense splice site or frame shift deletions. When the NF1 mutations and the BRAF mutations do co-occur they tend to co-occur with these Exxon 11 BRAF mutations which are weaker kinase activating mutations. Based on the relationship of these map kinase driver mutations we suggest a melanoma can be categorized into four genetic subtypes which include the BRAF hotspot mutations, the RAS hotspot mutations as we do identify a few H and KRAS hotspot mutations that are anti-correlated with the NRAS mutations. NF1 mutations which tend to be loss of function mutations in these triple wild type samples. So what's driving these triple wild type samples? So by incorporating other data platforms it becomes more clear. So again based on our sample size we're now able to do this type of analysis where we perform just a two analysis on the different subtypes. So we know mid-F which is a lineage specific oncogene is significantly amplified in melanoma but we only find this significant application in the BRAF hotspot mutant samples and same with the BRAF amplicon we only find that in the BRAF hotspot mutants. When we look at the NRAS hotspot mutants Andy Churniak identified a minimal common region that also includes NRAS and what you see in the triple wild type is you see this 4Q12 amplicon which is only found to be significant in the triple wild type and that contains KDR which is also known as VEGF2 and PDGRALTHA. We also observe significant applications in TURT Cycline D1 in this 12Q15 region often includes MDM2 and CDK4. So when we include these copy number alterations in the landscape we start to fill out what's driving this particular subtype of melanoma as we can see the different significant applications. We also have observed some cosmic mutations in known melanoma drivers such as GNAQ, GNA11 and KIT and this is just another representation of the distribution of these amplicons and you can see that they're enriched in the triple wild type. We also perform fusion analysis in a collaboration with the Harvard Medical School Brigham Women's and MD Anderson Cancer Genome Characterization Center where we perform low-pass whole genome sequencing. We also incorporated deep whole genome sequencing as well as RNA-seq using multiple collars. I won't go into this entire pipeline but on poster number 63 there's a description of this entire approach. We identified 221 potential drivers and I'm just going to highlight two. So these are the BRAF fusions that were identified. Recently in literature they've been estimated to occur in a relatively high fraction however we only see them occur or we only find two of them within all our samples. What's interesting is one of the fusions was only was found in the triple wild type and the other one we don't have overlapping exome data. So these are all map kinase driver mutations. Do they have similar signaling outputs? We looked at downstream signaling so phosphomech and interestingly we only saw comparatively elevated levels of phosphomech in the BRAF and NRAS mutant samples and only comparatively elevated levels of phospha work and the NRAS mutant samples and we're currently trying to decipher some of the signaling pathways involved in this through our integrative analysis. So what are some of the other interesting mutations we're identifying? So we identified this hotspot IDH1 mutations and unlike the GBM mutation it's not the R132H it's the R132C mutation which is caused by CDT transitions so presumably UV damage. The John Hopkins group performed clustering analysis of our methylation data and they've identified this high SIMP group and interestingly the IDH1 mutations clustered with this high SIMP group as well as ARID2. This high SIMP group also had less BRAF mutations and one of the issues with our clustering analysis is because we have metastatic samples from all different origins often there are reflection in terms of our clustering subgroups of the origin of where the tissue was procured however in this particular analysis we did not find that relationship. Another thing that's interesting about this is the GBM high SIMP group is often associated and IDH1 mutations are often associated with better survival and our analysis they're trending towards worse survival. One of the unique aspects of the TCGA project too is how we do our survival analysis because we are acquiring the metastatic sample and not the primary sample we are doing both overall survival as well as this TCGA survival and this is from when the sample is procured to the days to last follow up and death so this is how we're doing our survival analysis for the metastatic samples. When we performed the RNA-C group or RNA-C clustering analysis we identified three subgroups but as I mentioned before there are often a reflection of the sample where they were procured. So for example we identified this lymphocytic immune signature score or subgroup which was found to be enriched in the regional lymph nodes. We also identified a keratin or epithelial cluster which was enriched in the primary samples as well as the subgroup that was identified or that was distinguished by low mid-F and low mid-F signaling which is a lineage specific oncogene. Interestingly though the high lymphocytic group or immune group actually had the best survival and this is consistent with other studies that have been found in both many different tumor types as well as melanoma and this is in particular important given that the immune therapies have had a major impact for metastatic patients. So we dug deeper into this and we had our pathology team which was led by Richard Scullier. Actually look at the samples and score for lymphocytic infiltration so they're scored from low to high and you can see that there is distribution across the samples of low to high and even within the lymph nodes there are samples with low lymphocytic infiltration. And when you look at the survival you do see that the patients with samples that have high lymphocytic infiltration do better and even when you just consider this only the regional lymph node samples and again this is important in the context of the new immune checkpoint inhibitors targeted therapies for the immune system and melanoma. One of the other important findings that have recently come out of literature in terms of melanom genomics is these Tert promoter mutations that Matthew was asking about yesterday. So we do not capture the Tert promoter mutation obviously in our exome sequencing data. However the Harvard Brigham women's MD Anderson cancer center genome characterization center performed the PCR and Sanger sequencing on the over 100 samples in which they also did low pass whole genome sequencing and we observed that the Tert promoter mutations occur at a similar frequency that was reported in literature about 60 to 65% and that these two mutations were mutually exclusive. Going back to Matthew's question yesterday we found only that the C228 T mutation was associated with increased expression however again we're only looking at a hundred or so samples. Another preliminary analysis that we did observe was that the metastatic samples tended to have a higher fraction of samples that had this Tert promoter mutation. Here's just a high level overview of the pathways altered in melanoma is performed by the more Sloan Kettering group and Chris Sander and I know and what we're demonstrating here is in this in our study is this new MAPK driver melanoma NF1 found to be significantly mutated. These are the four genetic subtypes and as I pointed out through the course of the presentation we're finding certain driver events that are found in certain subtypes for example the triple wild type driven by MDM2 application, Cycline D1, CDK4, Tert and Kit. We also identified some significantly mutated genes that are associated with particular drivers that didn't get a chance to go into which include PP6C for example for the BRAF and NRAS mutants and this is becoming clear in our oncosine analysis which I don't have time to show you today. So in summary we've identified four genetically distinct melanoma subgroups which include the BRAF hotspot, RAS hotspot, NF1 loss of function and triple wild type. Integrating these platforms we show that they do not signal similarly through the MAPKINES pathway and so each of these mutations does have different MAPKINES signaling. We also identified the IDH1 hotspot mutation and the clusters with the high SIMP group and some of the analysis that unfortunately I wasn't able to show you today was the incorporation of the oncosine, the micro RNA clustering analysis as well as some of the primary versus metastatic comparative analysis that we've done as well as one of the main aspects of our project that we're kind of continuing with is the genetic determinants of this lymphocytic infiltration. So on that note I just want to thank everyone on the Manuscript Writing Committee that is involved in trying to put together this marker paper. I want to thank the Firehose or the Broad Firehose team that have helped with all the platform specific analysis and as of April this is all the authors contributing to the project. So thank you for your time and thank you to organizers for giving chance to speak today. We have time for a couple of questions. Yeah, the triple well type group is very interesting. Apparently the mutation rate is lower in that group. So are they associated with lower lymphocytic infiltration and that will infer that because most of these are the uni damage and the lung cancer they are generally neo antigens. Yes. So I'm wondering whether there's any correlation. The other question is for example that triple well type apparently is using different set of R.D.K.s and those are K.P.D.G.F. Those likely will respond to synitin able clients of R.D.K. drugs. Do we have any data for that? I don't have any data for that and I know people at MD Anderson and a few other groups are looking into that. In terms of your question about the lymphocytic infiltration and triple well type I do not think that there was an association in that case. I think it is due to the fact that they have a lower mutation rate but that is a good question and I'll look into that and get back to you. Regarding the triple well type since they don't seem to have a UV signature did you look for some other type of signatures in those samples and did you look also in the maybe in the germine to find out the repair genes would be affected. That's a great question. We are doing those analysis and we have looked at the mutation signatures that Michael Stratton identified and we also looked at chromothripsis and depending on the chromothripsis color you use we did see some enrichment in the triple well type samples. However these seem to be complex rearrangements not your classical chromothripsis. So that is one potential aspect of what's happening in these triple well type samples in terms of genetic signatures. Great thank you. Thanks very much. Next we'll hear from Rehan Nakbani from MD Anderson on the pan cancer proteomics landscape of the cancer genome atlas projects.