 Thanks very much, Mark. One of the barriers that was identified as an issue at the June meeting was institutional barriers, and the point was made several times in the course of that meeting that it was very important to have strong buy-in from the institution. And so we thought it would be valuable for us to hear some perspectives from some institutional leaders over why they thought this was an important effort for them to go forward with. And we're really delighted to have a couple that have made time to come speak with us today. And we'll start with Bill Evans from St. Jude's. Okay. Thanks, Rex. And I'm happy to be here and share with you the thinking of St. Jude. I guess by way of disclosure, I should tell you that I've been the director and CEO of St. Jude since 2004. But I've been an NIH-funded investigator in the field, like Dan Rodin, since the days when pharmacogenomics was called pharmacogenetics. And I still am engaged in running a lab and participating in that sort of research, which is actually the most enjoyable part of my job. But I'm here as an institutional CEO. And so let me just give you a brief bit of background and then sort of tell you where we are and where we're going next. I would point out, first of all, we started using genomics to decide therapy at St. Jude in 1984. And that's when we began to realize that somatic changes, major structural changes like chromosomal translocations or chromosome number were prognostic in children with acute lymphoblastic leukemia, the most common form of cancer in children. And so beginning in 84, we started using first ploidy and subsequently chromosomal translocations to assign patients to different levels of intensity for their ALL treatment, which could include a bone marrow transplant if they had some of the high-risk translocations, as indicated there, BCR-Able-MLF-AF4. And now we're actually looking at mutations in specific genes that can also be targets for therapy. Currently, that's done more in AML than ALL, but as I'll come to at the end, we're actually beginning to see where we can do that with ALL. Also, we began to investigate, this is sort of where I got involved, in the 80s inherited germline genome variation that could influence toxicity, effects of some of the cancer chemotherapy we're using. And we began in the 1990s, late 90s actually, to use TPMT genotype to decide doses of macapitopeurine. And I'm not sure if we have a pointer up here. But in work that Mary led, we also showed that not only the rare individual with two varying alleles of TPMT, but also heterozygates at the TPMT locus had a higher risk of drug-related toxicities. And so we began to adjust therapy based on that genetic polymorphism, which we now do routinely. We used to give all kids the same dose. They were quite different in their tolerance. Once we began to adjust doses, some kids are now treated with 10% of the standard dose if they have two varying alleles for TPMT. And they have the same toxicity profile then as the other kids, and we do that, and they have the same cure rate. In fact, perhaps they do even slightly better, but certainly no worse than their wild-type counterparts treated with, in many cases, 10 times the dose of that drug. And now there are guidelines on sort of how this is done. Mary's been leading this through the CPIC, and I'll come back to that in a minute. In terms of the approach we're currently taking to use this germline genetic polymorphism to individualize therapy. Now, eight years after those discoveries, we began to sort of do that and this issue of science, the genome issue, genomic medicine, issue in 2003 was pointing out that eight years later this still wasn't a routinely used diagnostic in the clinic. And even my friend Russ Altman is quoting here, quoted here, saying he thought this would be sort of the poster child for how to do this. It's so highly penetrant yet if you read this article and one in nature of the same weekend, it says the medical community remains skeptical. And I think in many ways they still do. I guess the other thing I should point out in the way of disclosure here, that there was a note in there that our institution had actually filed and obtained patents for these TPMT SNPs as diagnostics, and that allowed us to license them out in a non-exclusive way to several national reference labs that could offer this as a clinical diagnostic. A diagnostic was very expensive at the time, something we can now do for pennies, and that day took hundreds of dollars to get the test done under clear conditions. But just sort of take that with a grain of salt. A little bit of that royalty money came back to support my lab. I would submit, and this is a photo of me talking to a patient and a parent in 1999, and yes, I'm using a piece of film to have a conversation about a genetic test that we had done and that that was going to be used to change the treatment for this particular child in the capital period. Obviously, technology for genome interrogation and for identifying these associations has changed dramatically, but I would submit that translation continues to lag in terms of taking this to clinic, and that's why I applaud what's happening here at this and sort of sharing with you what we're trying to do to make that happen. So when Mary and I wrote this review, we put this sort of pie in the sky, the future technology would be some arrays, and on that we would not just interrogate TPMT or the somatic changes in leukemia cells and use that to guide therapy, but there would be a growing list of genetic polymorphisms and perhaps somatic changes that would drive that diagnostic that we would do upfront on every new patient with ALL or whatever the disease. So, you know, I think one other reality here before I tell you exactly what we're doing is that the academic system still continues to reward the discovery side of all this in terms of promotion and tenure committees and other kinds of academic currency which travel with faculty regardless of what institution you're at, but the translation and the recognition by institutions of the importance of translation is something that I think still remains very sort of institutional in nature and hopefully it's going to become more systemic. So what are we trying to do to translate these genomics more so more deeply into clinical practice at St. Jude? I would say that, you know, one thing is we put it in our strategic plan for the institution. So we had one in 2005. We just did a new one that we had approved last year. And as you can see, in addition to trying to push curates to 90% or higher for all childhood cancers, the fourth one on the list there says we want to be a model center for translating biomedical discoveries into innovative treatment strategies for children with cancer. I think it's important for the institution to sort of be public, at least internally if not externally, about this is a priority. And it sort of helps galvanize institutional decisions and thinking and participation of faculty. And then if you read deeper in that strategic plan, we also go forward and say to fully realize the promise of individualized medicine because it is so complex and not well understood and has to represent the integration of information and data from multiple sources that we need to give some support to our clinicians in real time in terms of evidence-based decision support tools that can affect and guide their clinical decisions in real time. So people always say, all right, you're an administrator. How much money do you spend? That really tells me your commitment. And I would argue that's not true, but I'm not even sure what the number is. However, I would say that of our $625 million annual operating budget, you know, we are predominantly a cancer center, so on the research piece it's at least half of our budget is invested in somewhere or another toward the discovery side of genomics as it relates to childhood cancer. The patient care side, it just, you know, how much of the hundreds of millions we spent on our electronic medical record and decision support is because of genomics versus other aspects of individualized medicine is sort of hard to quantitate, but it's not a trivial investment. And we're very happy to make that investment, whether it's $160 million or $220 million or even more, and that doesn't even count the high-seeks and the other equipment that we're investing in. This is just operating dollars. So here's an example. Now, as I said, we routinely used somatic genome variation to decide treatment of ALL. That continues to expand. And increasingly, and this is driven in part by the work being done, various labs at St. Jude, but also nationally, and the work led by Mary at St. Jude as part of the PGRN, is trying to identify additional inherited genome variations that we can translate to the clinic. And we think there are some that are of great utility, I won't call them actionable, of great utility. And so here's how we're trying to go about doing this. So once we identify a pharmacogenetic genotype that we think is clearly validated and of potential clinical utility, it goes on the problem list for the patient. So this is the screenshot from a problem list of our EMR, and it basically lists, as you can see there, TPMT deficiency as a problem in this particular patient. And then there is this behind-the-scenes decision support that now knows about that and will alert the physician should they begin to write orders on medications that could be affected by these genotypes. And another screenshot, what we want to do for all of our new ALLs, for example, they're all going on protocols, and most of the therapy given at St. Jude is on protocols, which is advantageous for what I'm talking about. But once they go on an ALL study, then it goes into the medical record to fire an alert that says you should go ahead and get a TPMT genotype on this patient, because in a few weeks they're going to be getting a thiopurine. And that would be useful, and you can even see at the bottom, it's pretty easy to just check the box and order the lab test to get a TPMT genotype on that patient. There's also a notice sent to the pharmacist, and we have this multi-disciplinary care at St. Jude that's been very extensively integrated that way for 50 years almost, that says, you know, a TPMT genotype doesn't appear to exist for this particular patient. You should know that they're now sort of in the queue headed toward therapy with the thiopurine, and you might want to talk about this on RALS to see if indeed you want to get this, or if in fact there's been an order issue, then this would alert the pharmacist to potentially intervene if they've somehow missed the alert that's come up in the medical record. Now if a physician sits down, a clinician sits down in front of the MR and begins to order this medication, then this warning would come up and says, this patient has an active entry on the problem list of TPMT deficiency, says this might be important. You can see at the bottom they have some choices. They can cancel the order they've entered and enter something different. They can alter the dose according to a consult that would have been provided if they requested or triggered electronically, and they can modify their medication, their prescription accordingly. So this sort of happens in real time. Now if they ordered this med and the patient were wild-type, they wouldn't get any alert whatsoever, because we're trying to avoid alert fatigue here. So it's only when we think it's something they need to perhaps take action on, that it's in the record. And then there is a consult that's generated through this process. They're standardized. The language has been sort of agreed upon in an effort led by Mary and others at St. Jude in terms of what the content of those consults that go into the medical record look like in terms of the recommendations that are being made. Now, the question is sort of how and when do we take one of these tests from the research side of the organization into the clinic? And this is actually from Mary's PGRN Power for Kids grant. Rochelle's here. A plug for PGRN. But part of the RFP said, we want there to be a translation piece to this science. We don't want you to just be thinking about discovery. And so this basically sort of points out how we go about trying to first discover and then replicate and then prioritize and subsequently validate through various strategies before ultimately taking one of these into the clinic in terms of a clinical integration step. And there's no one formula that any of us use or that I would advocate. And in fact, it might be different for different medications and different genetic traits. But there is a defined process. It sort of is an extension of the clinical research that's done here. And I think on the clinical integration side, we look at discoveries made anywhere in terms of them being valid and perhaps in diseases other than we are treating. But if we think they're relevant to using these meds in our patients, then we would formally consider the process of moving that into the electronic medical record into the support. Now, I show Mary here as the CPIC queen. I didn't get her permission for that. And I may pay for it later. But yeah, I will, Debbie, since, okay. There have been a number of published papers in the last year that represent this consensus view of whether a given genotype drug combination is ready for prime time. That is ready for clinical translation. And these are going to continue to grow. There's one impressed now. And they're being published in CP&T. But I think also importantly, they're being annotated on FarmGKB because this is going to continue to be a moving target as additional data come to make the association stronger or less strong in terms of clinical utility. But that certainly is one of the elements of information we use in making the translation. Now, I won't go through this in detail because many have already mentioned that we're all working against some gradient here in terms of trying to incorporate this into healthcare in this country because it's so fragmented. It's more sick care than it is prevention. The evidence, some think, is modest for many pharmacogenetic traits. And there is this genetic exceptionalism that these tests tend to have to overcome currently that other things in the clinic, whether it's proton beams or whatever, haven't had to overcome in a way of evidence. There's a lot of complexity and that's going to continue to grow. There's a lack of standardized computational decision support in all of medicine. We think there's a need for preemptive testing, particularly now, given you can do so many so cheaply in advance. Now, the advantage we have at St. Jude is many of these barriers actually aren't as operative at our place as there are a lot of other places. We pay for everything. We don't ask for any copays or deductibles. We provide all their healthcare, all their medications, inpatient, outpatient. Our EMR tracks their medication, their treatment for a lifetime, whether they were in our hospital or they were back in Southern California on continuation therapy. We're still tracking what's happening and the meds are getting, the outcomes are experiencing. Our patient care and research, like so many cancer centers, is extensively interwoven. We've been doing the multidisciplinary thing for a long time, and we've got millions of dollars invested. We use Cerner, but we've begun to individualize and customize our electronic medical record for providing this decision support. Now, the other thing, of course, that's happened is that we have now come to the point where we can do a lot of SNPs and a lot of genotypes on many drugs for relatively little money. And so it has become affordable for us to sort of do this, at least the DMET chip, up front on everybody and have that information available when needed to take action. And for the same money, we were doing a couple of genes, maybe even one a few years ago. We can now do all of these and have that available for patients. And that certainly is making this preemptive approach to genotyping and use feasible. So, again, there's a process here and I'm not going to go through it in detail, but we are now doing the DMET array up front under CLIA conditions. That's being done by Uli Brockel at the Medical College of Wisconsin. But eventually, I think it'll be imported under our own labs. We're using a very defined process to move this information from the research side to the clinical side, and I've told you about the decision support. We've begun to test drive this on a couple of genes, TPMT and SIP2D6. There is, anybody who knows, Mary knows, there is a very well-defined process. It's in place and I'm not going to walk you through this in detail or be in the slide. I've alluded to some of the components of that in terms of data supporting the clinical implementation, but also this process of having 1,900 SNPs on 225 genes for every patient. How many of those go in the medical record? Maybe not all of those are currently actionable. So the question becomes, how do you do that? When do you move them? And so the way it's structured and the way we're doing it at St. Jude is the CLIA D-MAT array is being run, and we're getting all the data, so 1,900 SNPs, 225 genes, but it sits on the research side of the firewall. And then when we think the evidence supports its translation across the firewall to the medical record, then we go through that process on the previous slide to make that decision and then ultimately migrate those 1 gene at a time and all their relevant SNPs going into the medical record and having the decision support that I told you about. Why are we doing this under a protocol? Which we are. It's currently every patient who has this done signs an informed consent on a protocol that has as its goal to migrate pharmacogenomics tests from the lab to the clinic. You could say, well, why isn't this just standard of care or does it test and do it? And, you know, there are a number of reasons why we felt it was important to do it under a protocol. Even though the assay is being done as it should under CLIA, it's still a very complicated process in deciding which of those tests should go in the record. We are thereby not putting everything in the record. We are withholding some information from a patient. And we want them to know we're going to do that and they have given us consent to do it. And we felt we needed consent for withholding some of the results and for dealing with the incidental findings in terms of other associations that have been linked, whether strong or not, with some of these genotypes like 2D6 or whatever. So we're sort of up front and laying this out to the patients. We would see this, we do see this as evolving to standard of care. Whether it takes us five years under this protocol to get there or 10, it remains to be seen. But the primary objective of this study is to estimate how many of our patients, given this diagnostic, are going to actually have high-risk genotypes, as we're calling them, which are actionable from a pharmacogenetic standpoint. And we have a number of secondary objectives that are sort of enhancing the tools and the process for prioritizing and assessing the attitudes and concerns of the clinical staff in using this. Now, any of you who would like to know more about this protocol can get it through the Internet website of St. Jude, shown here. And there's a video that we make available for patients or parents to look at that are prospective enrollees on this trial. And I know that Mary's shared this protocol in its entirety with some other investigators, so you could contact her directly if you'd like to see the entire protocol. I would say that after having enrolled, I think about 175 patients over the last several weeks or months on this trial, that about 15 percent, given just 2D6 and TPMT, about 15 percent have had what we refer to as high-risk genotypes that could be actionable. So I think that's, you know, that's about what we might have predicted, but certainly supports the use of this. It's not just a rare event for this to happen. A few other things that are done as part of this. We have an automated communication to the attending physician when a consult and a genotype is done and a consult is rendered so that they get separate notification electronically, not just having to go to the medical record to see it, but they would get it on their desktop or mobile device. And also, the parents can opt or patients can opt to receive a personal letter giving them the results of a genotype that we've now moved into the medical record. And all of them, essentially all of them have asked for this. And so we do provide that as well as this video. Here's an example of sort of one of these letters that will go out to the parents saying we did this genotype on your child and here's what we found, and this is sort of a blow-up of one element of that, but, you know, sort of tell them where they fit in the grand scheme of things of a population of patients. Are they extensive metabolizers or one of these other more rare genotypes for this particular polymorphism? And what does that mean in terms of our decisions as clinicians? And the process of putting together this protocol and on an ongoing basis, we have a St. Jude Family Advisory Council and they've been very engaged in sort of developing this trial and we ask them, do you want to know when we do this? And some say, you know, I don't even understand why you're telling me this, just do it. And others say, yeah, I'd like to know more about this and so they have the option for this to go one way or the other. Probably more patients and parents would say, I'm not sure why you're sitting down telling me all about this, just do it. It's another diagnostic test that you're telling me you need to do and you don't tell me you do a lot of other diagnostic tests and so it runs the gamut. Now sort of where do we go from here in terms of scaling up and we do have this big whole genome sequencing project and just sort of want to think beyond 1,900 SNPs and 225 genes to the future state which is not that far into the future and we have done almost 500 whole genomes in the last two years. Equal number of cancer genomes are running germline DNA across the full spectrum of pediatric cancers. These are not currently being done as diagnostic efforts, they are discovery efforts but I would tell you that our ability to generate whole genomes is like yours rapidly expanding so we're now generating two whole genomes every day and we will probably be doing twice that next year. So this is very feasible, I think all of us are seeing this and particularly those working in cancer areas and this is whole genome, this is not exome, transcriptome and if we sort of look at the landscape of genome variation that we're seeing these are somatic changes across these diagnoses you know there's a lot of difference from on the right the ones with high signals include osteosarcoma and some of the brain tumors to infant leukemias having relatively few Tier 1 mutations which are mis-sense mutations in coding regions a lot of structural variants. In a paper that's currently impressed will be out fairly soon just to illustrate in looking at a subset of T-Lenage leukemia compared to the garden variety T-Lenage leukemia so the early T-Progenitor TALL if we compare what we see in typical T-Lenage ALL where we see Notch 1, P10 and other well-known somatic changes in these genes other mutations or focal deletions we can see quite a different spectrum in early T-Progenitor where we're now seeing drug-targetable mutations in genes like flit-3 or jack kinases and so I think the future state is we're going to be wanting to do this on every patient to see whether they carry a flit-3 or a jack-2 or some other mutation that's driving their leukemia that could then drive a decision about treatment into the future and then the challenge will be how do we move 20,000 genes and 3 million variants for each patient across this firewall as we're now doing for 1,900 SNPs I would submit that sort of what we're learning now with the pharmacogenomics as I've just discussed it we'll inform how we do this on a much larger scale going forward but that is exactly where we're headed particularly in the cancer world so sort of to wrap up my thoughts and comments where to from here I do think over the next 10 years obviously it's going to become cheaper and cheaper to sequence a whole genome and our ability to analyze it is getting cheaper and better as well but that sort of becomes the right limiting step there's going to be a continuous expansion I would predict and valid pharmacogenomics traits so the pressure to do this is going to continue to increase these traits are going to I think become increasingly polygenic in nature and going to involve rare variants as common variants that collectively have to be integrated into more sophisticated models to make decisions treatment decisions based on inherited or acquired genome variation and this is a figure that Eric Shett showed Friday when he was at St. Jude in terms of thinking about networks and the integration and communication across networks that are ultimately driving phenotypes whether it's drug response or treatment risk or all the above I think our medical pharmacy and other health professionals are going to be better educated on this front but I don't think it's ever going to be sufficient for them to do this without support because it's a moving target and it's obviously very complex and we're just going to have to keep improving our systems and having experts be involved in this translation step and that's who's in the room so that's great and I do think it's going to continue to be a growing component of diagnosis and treatment decisions that's certainly going to be true in our place and I suspect most of yours as well so as I have alluded to there are lots of people involved Mary's a PI this study I've talked about this translation protocol if you will lots of players there and of course at St. Jude Pharmacogenetics I guess now Genomics has been a team sport for a long time and there are lots of people from our medical staff and PIs of trials to our pharmaceutical sciences group our informatics group our statistics folks a lot of students and postdocs so I'm going to stop there I don't know if you want to take questions now or after the next speaker I think we actually have time for a couple of questions and as the hands go up I'll actually start sort of the idea that you were able to make this an institutional strategic priority you're sort of a unique shop in that you're well known as a place that's a leader in terms of translating from research into clinical activity are there any lessons that might apply to places that are sort of less in that kind of a leadership role in terms of going to your boards and getting that kind of buy in and maybe as a related piece to that one of the things that strikes me is that cancer is sort of in a unique area in many of these cases nearly even at our place most cancer patients are on a protocol of one sort or another whereas that's not true in a lot of other areas how much does that benefit your institutional buy in? I think it benefits that at St. Jude it's you know almost all of our patients are on some protocol usually their primary treatment is driven by a protocol when I say usually 75-80% the reality is in adult oncology it's 5% so I think there's 5% of patients are on protocols the other 95% are out in best clinical management so I think adult oncology is not the same world what's that? not necessarily best okay I'm sorry yes what's considered most lucrative most lucrative clinical medicine according to an expert in adult oncology from an academic perspective so going to the boards you know this is a tough one because I sit on another hospital board at the university and I just can't get them to the boards like being leaders and being innovative and to say we offer personalized medicine or genomics medicine or whatever but they're not really often willing to put the resources in it to make it happen and if they can't get it reimbursed then they start saying well let's wait until it's more you know let's wait till we're forced to do it either by the other hospital in town doing it and we've got to or by lawsuits or by some other force so you know it's been very helpful that there's a research element a pharmacogenomics at St. Jude and so it's not viewed as sort of arbitrary but being driven off of science and our board's been very receptive to that approach whether it's the way we treat cancer or the way we move genetic test into the medical record but you know I would say that the translation piece is not funded by NIH the discovery piece is it's also funded to some extent by donations because we can never get a hundred percent on the dollar for my NIH grants right and so Saturday we were at the St. Jude Marathon which raised four million dollars yesterday I was in the hospital touring Taylor Swift so there's a lot of other efforts going on to bring the resources to bear to be able to do this so I have a couple of questions I'll ask some sequentially since they're not directly related the first is you're in a unique position also because you are also the payer so you're really integrated on the cost side so I'm curious if you've done formal return on investment analysis on what this impact has had for your cost structure yeah I haven't done it on this project per se but we did it on how we treat ALL and we compared what we spent at St. Jude to the cooperative group which is the other 200 institutions in the US and when we did this we were spending $150,000 per year per average and the group was spending 50 so I told the board we are spending three times more dollars per patient and I said but if you look at the number of papers published per patient treated from St. Jude versus everybody else we were publishing 20 times more papers we were spending three times more dollars so to me it was a seven fold favorable return on investment that's why you're in grace it all depends on the metric of success well it does and you know I'm one of the few CEOs that I used to be could lose $350 million a year and keep his or her job now that's pretty common on Wall Street but we were there early in that paradigm in fact you wouldn't qualify for your top bonus if you only lost $350 million a year the other good thing about St. Jude there's no compensation incentives based on the bottom line it's all based on new knowledge we do two things state-of-the-art treatment and generate new knowledge and so the return on that investment is more publications more knowledge so the second question then relates to that and it seems from your presentation that you give providers the opportunity to choose am I going to follow this instruction or not follow this instruction so the question is are you tracking those decisions and are you tracking the patient outcomes based on whether or not say a TPMT protocol is followed versus not followed yeah we definitely track that now the good news for us is all the medical staff work for us so there's no one coming in out of private practice and opting not to do this we definitely give our physicians a choice I'm glad that came through but if they don't make the right choice we give them a lot of coaching how's that sound they're terrified when Mary shows up in their office yes we have published data on lowering toxicity of thiopurines it's been replicated around the world it's very solid we have other data on the 2D6 encoding lack of response it's coming from other institutions but we share that with them and say this body of evidence has convinced us that patients will have better outcomes now ultimately we're not going to have those that did and those that didn't get pharmacogenomics at St. Jude it's just not a comparison I'm going to make internally because we're going to make sure all the kids are going to be useful in the way of a genetics test but when we look at our outcomes versus the rest of the world and they're better we won't know for sure how much that contributed to it but it might be part of it I think we need to move along but Eric had one quick question that's one quick question you're a statistic about you're sequencing two whole genomes a day that's you guys or that's your partnership with Washington University Center they're doing all the validation work they're doing all the high throughput sequencing of whole genomes and sending us the data that we're both doing it's a completely joint effort paid for by St. Jude we would love for NIH to help if you would thank you very much Bill that was great