 Okay. The next report we have for council is on cancer genome atlas. The atlas project was begun, I believe, in 2005, it's a massive multi-omic study that had an overarching goal to define the molecular genetic changes associated with a whole slew of different cancer types. The NHGRI production sequencing centers were tremendously involved in this project and many aspects of the project, but most notably in the data production work. So we thought it was appropriate to bring a presentation to the council. I think the production work stopped in 2015. Carolyn can clean up any of my mistakes that I make in the intro. Some of the analysis work is continuing, so we thought it was an appropriate time to do this. And what better person than our own Carolyn Hutter, who probably gave a couple of years of her life to the coordination and management of this project when she worked at NCI? So Carolyn? Thank you. So yeah, I'm going to be giving an overview of the cancer genome atlas, a decade of discovery. I guess we're actually into the second decade, was it being 2017, but so the cancer genome atlas or TCGA, as I'll call it, was a joint project between the National Cancer Institute and the National Human Genome Research Institute. The mission, as I have outlined here, was to really take a comprehensive look to accelerate our understanding of the molecular basis of cancer, taking a number of genomic analysis technologies, including large-scale genome sequencing. This is a picture down here of me, Carolyn, with my co-workers at NCI Tiffany and Audrey, and we cheated with Liz Glanders, showing our base pair pride in TCGA order at the NHGRI symposium exhibit downtown. So TCGA came about in this comprehensive way to look at over 10,000, the number actually wound up, the goal was 10,000, the final number actually wound up being 11,000 tumor normal tissue pairs from participants with 33 different types of cancers, including 10 rare cancers with seven different data types, and I'll go into what those data types are in a couple of slides. The start of the project, and we're sort of this decade of discovery, really came out of these two seminal activities in 2005. One was a report to the National Cancer Advisory Board, and the second was a meeting about a comprehensive genomic analysis of cancer, and at that time, what came out of this was the development of the pilot phase of TCGA, and the pilot phase was designed as a collaboration again between NHGRI and NCI to test the feasibility of taking a genomic approach to cancers starting with three cancers, brain, ovarian, and why did I just space out on this? Lung, thank you, lung cancer, and really the complexity, the idea was that the complexity of cancer really required this comprehensive look at the genomic alterations to really think about understanding cancer and thinking about medical solutions and new medical approaches to cancer. I always like to sort of put this 2006 pilot phase start in perspective, and one of the ways to do this is to look at where we were in the idea of genome sequencing and the cost of the genome. So this is the iconic NHGRI cost per genome chart showing when this project started the cost to do whole genome sequencing was about $10 million, and the reason I do this is I think it shows the audacity and sort of the vision of this project. I also think it's important because I get a lot of people coming with these very 2016 glasses or 2017 glasses and being like, why didn't you guys hold genome sequence everybody in TCGA? And it's like, well, yeah, if we started TCGA today, we would have done such things. We didn't start TCGA today. We started TCGA at a time where we weren't even whole exome sequencing anybody. We were starting with a candidate gene approach and the project really evolved with the technology and a major part of that involving was three years later after the end of the sort of pilot phase with the expansion phase and the expansion phase was brought about by two three things. One was the success that was found in the pilot, two was sort of the evolving technology and the recognition that there was more that could be done. And finally three was as many of us know 2009, 2010 was when the R.S. stimulus funding happened and that actually allowed a fair amount of money to be brought in to really jump start this project from a pilot phase to a full phase, which at that point had a goal of doing 10,000 tumor normal pairs in 20 different cancers. And the thought at that time is the expansion phase of TCGA was really going to bring a quantum leap in our understanding of cancer and by having this molecular characterization on that number of individuals and that number of different cancer types we could really change how we think about and understand the molecular basis of cancer. As Rudy noted, the production for the project ended in 2015. At that point we sort of tied up the last but there was a little bit that went into like January, February of 2016, but we did a pretty good job of hitting that 2015 deadline for production. But one of the things to think about is, and this is a quote from John Weinstein, our current steering committee chair, is that although the project is concluding, it's really just beginning. And I think that's an important part of thinking about TCGA not just as a production thing but as a resource that really can be mined and used in a lot of different ways and is continuing to impact how we understand and think about cancer. As I mentioned, there are seven data types that were collected on all of the individuals in TCGA, most relevant to the sort of NHGRI perspective and what our sequencing centers involved in was DNA sequence. In the end we wound up with whole exome sequencing on all of the cases and whole genome sequencing on about 10 percent. So about a thousand subjects that had both whole exome and whole genome sequencing. In addition to the DNA sequence, there was RNA sequencing which also allowed for RNA expression. Copy number alterations were measured from a SNPCHIP, DNA methylation, microRNA expression, functional protein assays, and clinical data. Now the clinical data in TCGA is admittedly limited and part of what's happening now in NCI efforts is to use TCGA pipelines and approaches in cases that have richer clinical data. But there was a minimal amount of clinical data that was collected in the context of TCGA. And all together there's over 2.5 petabytes of data in this project. It was really also in addition to pushing the technologies, it really pushed a lot in the idea of data science. As I noted in part of the reason that we talk about the project, why the project is going to have a continuing sort of legacy post-production is a key goal of the project from the start was to have the data publicly and broadly available to the research community while obviously protecting patient privacy in the case of controlled access data. And there's both controlled access and open access data and data types within the TCGA project. And to give an idea of the sort of scale to which this data has been used, focusing only on the controlled access data, what people have to go through DbGaP to get access to, there's been over 2,700 approved data access requests for the data in the 10-year history of the project with many, many more people not even going through that process and using the open access aspects of the project. Over the history of the project, this data has sort of lived in different places. In the early part of the project before 2012, controlled access data was in DbGaP and then open access data was available through the portal. In 2012, while DbGaP as an authorization and authentication method has remained for how you access TCGA data, the actual BAM data, the sequence level data for both DNA and RNA moved to the cancer genomics hub or CG hub with the open access data stain in the data portal. And in 2016, NCI developed the genomic data commons or the GDC, which now is a single source place to find all of the TCGA data along with other genomic data that's generated through NCI. And so the genomics data commons, the link for which is down here, is the go-to place if you're interested in accessing and using TCGA data. I don't have the slide. I feel after Rex's, I realize I probably should have did it, of how many publications have used TCGA data. I chose instead just to go to the NCI genomics Twitter feed, which is sort of the follow-up to the TCGA updates Twitter feed, and pull some recent examples. These are all papers not done by the TCGA network, but by people outside of the network accessing and using the data. And that really is what we view as one of the greatest successes of this project, anywhere from sort of looking at drugable or potentially drugable variants found in different genomes to looking at how different types of cancer compare to other diseases. There's a variety of different uses. And one of the things that we talk about that really is amazing to see is the situations where people are using the TCGA data for things that nobody ever would have thought of, and really the ability to come in and do those types of work with the data is very rewarding. In addition to the data as a community resource, there has also been network-specific analysis done by the funded investigators within the TCGA network. Over time, this also included some people who sort of, I think, volunteered their time. But for the most part, we developed 32 working groups defined by tumor type. Those of you who've been paying attention might have said before you said there were 33 cancers and now you're down to 32 working groups. And basically, we just combined the colon and the rectal together. And colorectal actually only was a single working group. For that cancer type. These working groups each took the number of samples. Most cancers had, most of the common cancers wound up with between 200 and 500 cases. Breast cancer, there was 1,000 for the rare cancers. And again, we identified 10 rare cancer types. There was more in the order of about 50 to 100 subjects. So the tumor normal samples, the seven data types, and the working groups included people from the genome sequencing centers, the NHGi funded centers, the genomic characterization centers, which were collecting these other data types, the genomic data analysis centers, as well as some of the tissue providers and experts in that particular cancer, who did comprehensive characterization of the cancer genome across each of these different cancers. These all get published in what we sort of call marker papers or comprehensive analyses focused on a single cancer type. We've actually now published 28 of these marker papers, eight of them coming out in the last year. Esophageal cancer, pancreas and uveal melanoma being examples of papers that were published this summer. And the final four have actually all been submitted. And so they're in different stages of the review process. And we do hope to have all four of those, hopefully, come out this year sort of finalizing that network organized analysis on a tumor-by-tumor basis. Key results and findings that have come from these papers, I'm not going to summarize 28 papers worth, but we sort of categorize them into three different areas. One is just the profound understanding about the molecular basis of cancer, the genomic underpinnings of cancer. What are the significantly mutated genes? What patterns of expression do we see? What pathways are heavily involved in different cancers? Interesting findings, for example, in head and neck was really seeing a difference between the genomic profile in people who have HPV versus not. How that viral cause of the cancer leads to a completely different molecular profile when you look across all the different types. I think to me what's been most interesting is this tumor subtypes and recognition of molecularly characterizing cancer and looking at how we classify and think about subtypes and ways of cancer beyond. Some of which complement and some of which build on histological or other subtypes of cancer. So stomach cancer, for example, they were able to identify four separate subtypes, each of which have different sort of prognostic and therapeutic implications. And then the paper this summer focused on esophageal cancer, actually found that the squamous cell, esophageal, and the adenocarcinoma esophageal are just completely distinct from one another. And our squamous cell is more similar to other squamous cell, for example, in head and lung than that other cancer in the esophageal. And then the esophageal adenocarcinoma is very similar to a chromosome instable form of stomach cancer. And so that type of classification and changing, especially when these types have different prognostic impacts, change how you might approach or treat these. They also really start to change the way we think about informing and designing clinical trials. How do you bucket and bring together different types of cancer? Or not in the structure of a trial leading to sort of more modern umbrella and bucket style clinical trials. There's also obviously therapeutic targets coming out of TCGA. As we start to identify genomic characteristics of targets that work with currently available therapies or also might inform new drug development. If we start to find certain genes or certain pathways are highly hit or highly altered within a specific cancer type, that leads to reasons to think about drug development that would target those pathways in those cancers. So for example, some of the work coming out of the lung squamous cell carcinoma really has informed the current lung map trial, which is working in NCI and looking at specific genomic changes in their tumor. And that's just one example of many sort of precision medicine trials that are happening, many of which are informed by findings from TCGA and other similar studies. But what's been a major focus as well within TCGA at the network level. And again, this is also being done by people who access the data. There's been a fair amount of sort of cross cut cancer projects or what we in TCGA call pan cancer projects, where you look across the tumor type. So not just the single one cancer at a time through 32 working groups, but looking across. And so the first set of these, the pancan 12, as we called it, was published in fall of 2013, focusing on the 12 cancer types that had a sort of significant amount of data produced at that time. And so it was, oops, I don't know why I just went the wrong way. Sorry for the technical difficulties. So looking across 12 different cancer types and this counting for this project, we actually separated colo and rectal cancer. Some of these abbreviations may not mean a lot to you. When you work with TCGA enough, you sort of memorize these cancer codes. But the main point, and this is focusing on this slide, just from the sequencing data, what types of patterns and things did came from looking at the sequence data across 12 different cancer types. And so there were key findings, some of which were known, but some of which were highlighted through this analysis about the mutation rate, that different cancer types actually have different mutation rates or a different frequency of mutations within individuals with that cancer type. Some of this correlates with environmental exposures. So lung cancer and smoking actually tends to have a higher cancer rate than others. There's also mutational spectrums or mutational signatures that come out of this, where you can actually tie a mutational signature to a environmental exposure. So people who have their cancer because of smoking or have their cancer because of different exposures can actually have a different pattern in the spectrum of mutations or the types of mutations that are occurring in those cancers. We also saw a fair amount about mutated genes. What are the genes that are frequently mutated in different cancers? What are shared in different between cancers when you look at that? What is the mutational relationship? In some cases you'll see that there's mutually exclusive cancer mutations. So if you have a specific mutation, you're not likely to have the other. And how do those patterns differ and remain the same across cancer types? And then getting also into clinical features in clonal architecture. But some of what's even more interesting than just looking at the sequence data alone is when you do an integrative analysis bringing all of the different data types together and create subgroups of these different cancer types. And so not asking about what are the subgroups within stomach cancer, but saying, not looking at where this cancer came from, let's just take the data and have it cluster cancers into different groups. And what do we see when we do that? And in some cases, you get what you expect, the ovarian cancer, all sort of clustered together as a distinct group. But in some cases, you had differences. So basal breast cancer and luminal breast cancer were as different from one another as they were from other cancers. And bladder cancer wound up being interesting in that some bladder cancer just more sort of clustered with lung and headed next squamous cancer. And as we move to, including esophageal now, the esophageal squamous is going to cluster in the same group. Whereas some of it clustered with adeno and some of it was a separate group. And so even at the sort of anatomical level, there's similarities and differences when you say, what do I see about a cancer looking at it genomically versus what do I see using traditional classification? And again, I think this really impacts the way we think about and characterize cancer. So we're moving now from pancan 12 to what's in progress, which is the pancan atlas, which is what we're sort of considering the TCGA capstone project, looking across the full spectrum. It'll probably be closer to 10,000 cases than 11,000 once we QC and see who has not missing data across the whole thing. Doing a full data analysis and really coming in and not just saying what's the analysis, but how do we do a final summary about standards for data quality, data reproducibility. What types of batch corrections do we need to do as we move forward and decide to analyze a project that was pretty uniform. But if you think about again, being collected over a 10 year span with the changes that come into place with that. What types of re-analysis do we have to do? And then how do we also think about really analyzing this size and scope of a data set? Each individual cancer was about 500 cases to 1,000 and now all of a sudden we're talking about 10,000 cases. From an NHGRI project perspective, one of the things that we've been helping coordinate and Heidi Sophia and Melpy Kasopi who are in the room really have been leading this on our side is what we call MC3. Or multi-center mutation calling for multiple cancers. Within TCGA we had what was MC2 which was sort of not just taking a single caller for somatic variation but taking multiple callers. And recognizing you get a better data set with better characteristics in terms of positive and negative predictive values if you use a consensus caller rather than a single caller. There's not one single one that sort of wins out. And so originally in TCGA each of these 32 groups had gone through and done sequencing. And for the last set maybe start, I don't remember when MC2 started. We had the multi-center calling, but it wasn't a completely consistent set of calls across all of it. And so in order to be able to do this type of analysis we recognize we need to go back and do multi-center calling across all 10,000 sort of at once. And this involved 500 terabytes of BAM exomes to come up with this uniform mutation calling across all of this data with these multiple variant calling tools and appropriate QC filtering and annotation, etc. And so part of what was able to make this work was really developing a reproducible tools and workflows project and coming in and having these put into Docker containers using workflow specifications. Leveraging clouds and clusters. We actually did leverage DNA Nexus as well for they did a large amount of this calling using open source callers that we provided to just have a place to do that level of computing. And in the end what we wound up with is a consistent consensus call set across all 10,000 cases with multiple callers that's available publicly and is also really going to feed into being the variant calls that are used, that's being used for all of these pancan analyses. And the themes for this final pancan analysis, the papers are being sort of organized into three themes. One is oncogenic processes, which is really looking at what are the key driver mutations, what are the key patterns in terms of genomic alterations that we see when we look across the cancer. A cell of origin, which is not the best name in the end of what we're doing, but that's where we started doing, instead of just doing stomach and esophageal separate, what happens if we go looking across the entire GI tract? What happens if we look at reproductively associated cancers? What happens if we look at all the squamous cell together? Sort of reorganizing the focus and seeing what we learn from that. And then a series of pathway papers that are really looking at specific pathways that are known to play important roles in cancer and looking across the cancer sets and seeing what do we see similar and different in different cancer types focusing on specific pathways. These papers are in various stages of final drafts through submission and we're hoping that there'll be something that'll be published within the next year, knock on wood. While we've been doing this pancan atlas sort of capstone within TCGA, we've also been involved in a separate pancancer project, which is the pancancer analysis of whole genomes or PCOG, as you'll hear it called. And PCOG is a collaboration between TCGA and ICGC, which is the International Cancer Genomes Consortium. And through this project in PCOG, they actually got whole genome sequencing on 2800 cancer. So a fraction of these, about 900 of them came from TCGA with the remaining set coming internationally from lots of different countries around the world. And they brought all of this whole genome sequence data together. Again, similar to TCGA, took a distributed computing, large scale approach to doing standardized variant calling. This picture shows the different computational centers that were used within the PCOG project looking at this. In the end, 2,583 whole genome sequence donors that sort of passed the final QC and sort of made it to the final analysis where they had over 50 million somatic variants identified across all of those different cancers. The final variant calling, even with this large amount of sort of donated and used compute, took 23 months to actually do the full set of whole genome variant calling across multiple collars in this standardized format. And one of the things that was interesting in the context of this project is you didn't just have technical issues. You also ran into a number of both ethical and also political issues in terms of where can the data even live? Like who's willing to have their data go to Amazon if it's going to be in Ireland versus if it might be, you can't necessarily control in some of these cloud systems what country the data is in. And different countries have different regulations surrounding that that led to additional complexities that the group was able to work through and go. And so similar to the TCGA effort, this is now in a final stages of submission of papers. And so that'll be a second complement or suite of papers with a real focus. In this case, TCGA, we're focusing only on the whole exome sequencing. And this is really, but again, the seven data types. And this is really focused more specifically around what you see through whole genome sequencing. So I think as we think of coming to the end of this 10 years and into our second decade, the impact of TCGA comes in a different way, in a bunch of different ways. One is I talked about the creation just of this comprehensive data resource that is publicly available and that people are able to pull from and mine and use I think will continue to have a major impact, not just in cancer genomics, but in genomics more broadly. I think that TCGA has had a lot of sort of forward-looking pushing with data sharing and the move from DBGAP to CG Hub to the GDC really has sort of along the way also pushed policy and pushed a lot of questions about how do we do this. You know, the first trusted partners, the NCI Cloud Pilots, all of these things are really coming about as part of TCGA. There's been transformational scientific advances in terms of our understanding of the genomic and molecular characterization of cancer. And there's also been a lot of transformation and innovation in the pipeline and the approaches. The way that you even go about doing this type of work is really been informed by TCGA. And NCI now is continuing to do a fair amount of genomic characterization. As I said in studies, precision medicine trials, epi studies, other places that have more clinical data using the pipelines and approaches that were really sort of developed and honed through TCGA and not sort of reinventing that wheel. Just a final thing, there's a little bit of a formatting issue on this computer with this one. But to save the date, there's going to be a cell symposia about the TCGA legacy here in Washington, D.C., a little over a year from now. So if you're interested, please keep your eye out for that and that announcement. We're really sort of seeing this as a, you know, a lot of these pan cancer and PCOG papers that I've discussed will be out at that time and really a good time to sort of come together as a community and really take a look at what the legacy and impact of TCGA is. And on that note, I'll just end on my acknowledgment slide. Again, I already mentioned Heidi and Melpy are sort of the current project team, TCGA project team members with me. Past TCGA project team members, I certainly would be remiss if I didn't call out Brad Ozenberger, who was the NHGRI lead. And then on the NCI side, currently JC, Jean-Claude's inclusion and his team there and the TCGA DAC, Vivian Odoeng and Jeff, who sort of managed those almost 3,000 data access requests that I mentioned. So, thank you. Questions for Carol? Carol? So thank you for that. Have a great day. That was great. So do you think that the uptake of TCGA data by investigators outside of the TCGA consortium, do you think, is there a sense of how much that was fostered by simply making the data available or was it by making interfaces to the data available? Like CBIOPORTAL makes the TCGA data accessible in a analysis sort of way, in summary way that would be difficult to repeat if you just downloaded all of the TCGA data set. So do you have a sense of what led to the uptake of the data? Yeah, I mean, I don't know that we've studied that as well as we could have in a scientific way, but I definitely think that the ways, the investment that's been put into making the data not just available, but really tools and integrated. And I think CBIOPORTAL is a good example. A lot of the work that's coming out of the GDC, they just started a new, it's called DAVE. It's like data visualization. I'm going to get the acronym wrong. I have been really critical because the number of people who could just download the full set of TCGA data and work with it is not that large. Even with the MC3, people who could work with the multi-center mutation calling from a single cancer set were like, I can't work with this new large thing. And so sort of processing some of that in and making it available, even not just the people outside of the network. There's a large number of the people in the network who actually only work with the products from the broad pipelines, from CBIOPORTAL, et cetera. And so I think that has been a really critical part of what's made the data as useful as it is. And I think it's changing. And now there's some clinical researchers who are like, oh yeah, I can work with a small math file. I can do that because it's provided in ways that that can be done. Rafael? Thanks for the presentation. So I imagine that you learned a lot, the whole team learned a lot from organizing this apparently very, very complex project. Is there a way for you to pass on this knowledge to future projects? Yeah, so we try to do that in some different ways. JC and a couple of people from his group actually wrote a book. You can find it where that talks about the management and the types of things that go into that. And then I think it's also, but how to really codify that and do that, I think is a challenge. And it's one of the things that we try, for example, at NHGRI and NCI, also to sort of use the knowledge that we've gained and sort of apply it in other projects. I don't know if you have suggestions or thoughts about specific things that we could be doing in that area. Right now I don't, but I would think you would have more suggestions than me. Yeah. Was it on that specific topic, Eric, or? I was, okay, I think we can continue. So I have a question along the very same lines. When I think of TCGA, I think that the path from discovery to translation was amazingly fast and fairly successful. Do you think that's because cancer is a particular disease because of the role of somatic variation? Or are there lessons learned to push translation that NHGRI, for example, you think of the common disease program, the CMGs, and emerge, what can we learn about sort of greasing that pipeline that works well from discovery to translation that could benefit all of us? Yeah, I mean I think it was uniquely suited in that somatically, once cancer is a genomic disease, right, and there's so much of the way that you can remove a tumor and look at it and study it and really understand that genetic characterization that's more difficult to do in other diseases. You're not just looking at risk or predisposition to the disease, you're actually looking at the manifestation of the genomic manifestation of the disease. So for example, in the Centers for Common Disease Genomics we're focusing on the germline risk factors. We're not able to sort of have a cardiovascular equivalent of let's look at the somatic profile of the tumor at least currently. So I think that's helped a lot in terms of sort of the movement to translation because you're then identifying these genomic alterations that you could come and therapeutically target. I think there's also, some of this is also tied very closely to the existing cancer clinical trial network. And I think being able to bring TCGA in line with these precision medicine trials and I think thinking about how to really find primed either through things we're doing like Ignite or other projects for some of the other areas we're working in to say how do we then make that connection to clinical trials and clinical applications is a key part of it because it wouldn't have happened from TCGA alone, right? It had to be able to be carried over to that side. And I do think NCI is well-positioned to help with that. Yeah, I guess I would make a related comment. Thank you, that was really lovely. And I do think that as you just said, right? The cancers that were picked, particularly the early group were all cancers with very poor outcomes, right? And so there was a huge need for better treatment, whereas frankly the cardiologists have done a bit better than we have. But you didn't comment at all on the germline. So I did want to comment that there are, because I know members of my group are involved. There's now a very active germline committee, PAN Cancer Germline Committee, that is now heavily looking into the match normal data and trying to get new insights into cancer susceptibility. So I think that's another area where we're gonna continue past this original TCGA effort. Yeah, no, that's true. And Deb's one of the people who's definitely leading that, Deb Ritter. And that's happening in both of the PAN-CAN projects. So this focus on germline, both in the TCGA, PAN-CAN Atlas, and then also in the ICGCP cog. The ICGCP cog effort gets a little hurt by the fact that they have a much smaller sample size when you go to germline, the sample size difference, and then the whole genome focus. But both of those groups, it's also, I mean, Lisa Brooks has for the longest time been like, you have all this germline data and variants in it, and yes, they do have cancer, but the data is there. And I agree that's yet another area where having the data available and the uses, it wasn't designed to be a project that worked at germline at all, but now the recognition that we have 10,000 whole exome with matched normal, there are some questions about how normal is the normal, but it still is an informative data set to use in those types of ways. Dan, last question. So I had to ask Terry this question because I was embarrassed to ask it in public first, but my question to her was, when was it recognized that cancer is a disease of the genome? This is like past history, but when you started, was that? Yes, it was. Yeah, I'll recognize. That was well recognized. I don't know. I mean, Sharon, when would you, I feel like it's been known for. Well, as you can say for 50 years, but the other reason? I think it's, because of the classic aberrations, like the translocations and end-mic amplification, I think it's been well recognized, it's been well recognized that we did not know the major drivers of the most deadly cancers. I think that's where TCGA really led off with, okay, we know them for these rare tumors, but we really don't know them for the common tumors that are causing a high rate of cancer death. Right, but even at like a karyotype level, it's been well known for quite a while. I would have to, I shouldn't, I can't say for sure when it started to be known. I was just trying to sort of go back to your comment about when you started and the genome was $10 million. Right. It was a bit of an act of faith to sort of walk off the edge of a, it's like in that Indiana Jones thing, you walk off the edge of the cliff and you hope that there's something underneath you. There was, for example, already the EGFR result in lung cancer, which I think was a very substantial one in convincing people that TCGA had knew. No, it wasn't like this. Yeah, I mean, I think there was enough to sort of a Veeves point, there was enough evidence to know that something would be found from coming in and especially starting with the pilot to sort of say, we're gonna take a look at this and even the three pilot cancers provided enough, by the time it went to the full-scale 2009 acceleration, it was very clear that we were gonna have a much better understanding from. My view as an outsider was that there were the pre-pilots, which were things that existed before TCGA were the proof of concept. The pilot was about how to actually make it work in a kind of pragmatic way. And then after that. And to really scale it. You scale it, you work within the context of the technology. And I think one of the challenges at the time is that the people who were working on planning for TCGA, they knew that they were on a curve and the curve was moving. And a lot of the, I remember people having those conversations on whether to switch from this to that. There were arrays to RNA-seq for the RNA side. There were arrays to all exome sequencing for the snipping side because the techniques kept moving and you could always wait for the better one to come and it would come. And so agreeing to continue while techniques were changing is one, I think actually of the good demonstrations of TCGA is that you can do it without becoming paralyzed. But I think before this turns into a love fest, I can remember a Sunday Times front page article that was very critical of TCGA. And so there was a risk involved and it worked out. Steve Ellidge wrote an editorial in Science that was highly critical. And that was what this New York Times, that the work should go towards known pathways. I mean, so there were definitely. Yeah, and another area of critique was actually against the size of the investment and the translated into what's the R01 equivalent of the TCGA project and if you compare those, where are you getting? I think that criticism has really sort of gone away in the last couple of years as the projects really gelled together a lot more. That's not one you hear as much anymore but it certainly was there four, five years ago as well. Okay, thank you. Thank you, Karla.