 the 10th in the current topic series. Our speaker this morning is Dr. Howard McLeod, who recently moved to the Moffitt Cancer Center in Tampa. He's a medical director at the DeBartolo Family Personalized Medicine Institute, as well as a senior member of the Division of Population Sciences. Dr. McLeod's a leader in the field of pharmacogenomics, a relatively new discipline that explores how genetic information influences our response to drugs. His research has already had several effects on FDA policies. For example, he and others have shown that genetic variants play a role in patients' responses to warfarin, a blood thinner prescribed to more than 2 million people in the United States. Based on these analyses, the FDA issued new dosing guidelines based on the genotyping of two genes. As the new medical director of the Personalized Medical Institute, Medicine Institute, Dr. McLeod will be involved with the Moffitt's Total Cancer Care Study to create and share targeted cancer treatments that will improve patient outcomes. During this morning's lecture, Dr. McLeod will be expanding on both these stories, as well as telling us about other developments in pharmacogenomics. Please join me in welcoming Dr. McLeod to the NIH campus this morning. Thank you. It's a pleasure to be back and to update you on what's happening. These things happen every two years in this course, and it's always surprising how much has happened and how little has happened in two years. And that's true with all disciplines. There's things that we're still talking about today that were discovered decades ago, and there are things that have moved on to the point where we don't even talk about it anymore because they've become routine. And so that's certainly true in pharmacogenomics. I'm going to talk a little bit about some of the emerging trends in the field, some of the ways that we've been thinking about how to advance the discipline, and then also some of the ways we're trying to make sure that it's not just the rich countries or the, in our case, the formerly rich countries that are benefiting from the genomic advances that are happening. Now, I like to start pretty much every presentation that I give, not with this, although I will put that. I do, I'm on the board of directors of a small company down in RTP that is a pharmacist services company, but is unrelated to the topics that I'm speaking about today. I'd like to start with this particular quote, and that is, a surgeon who uses the wrong side of the scalp will cut their own fingers and not the patient. If the same applied to drugs, they would have been investigated very carefully a long time ago. This quote is supposedly from 1849. It's a journal that I don't read. Of the drugs that are approved by the US FDA in 2014, or since the beginning of time until 2014, there are none of them that we really know the mechanism of action. We know something, we call them something, we call them a cycle oxygenase two inhibitor or a topoisomerase one inhibitor, but COX-2 inhibitors have activity in COX-2 knockout mice. There's something else going on, and so that's true with all of the medicines that are out there. We know something about them, but not a lot about them, and so there's a lot of advances still in terms of discovering what the genes are that regulate the effect of these drugs, how can we use that information to guide therapy, or how can we use that information to even just counsel the patients better in terms of what to expect in cases where we don't have a lot of alternatives. Thankfully, in modern times, for most diseases, and I know this is the NIH campus, and so there are a lot of extremely rare diseases that are seen in this building, but for most diseases, certainly the common diseases, there are many active treatments that are available for use in modern times. If you take an extreme example, for the treatment of high blood pressure, the US FDA has approved more than 100 drugs or drug combinations for the treatment of high blood pressure, and so when you sit down with a patient and try to decide which medicine should we give this particular person, how do you choose? Well, you choose the one you know how to spell is almost the way we do it. There really isn't a lot of science that goes into it. It's more clinician familiarity and sometimes cost, sometimes other elements, and often it's trial and error. We'll try a beta blocker, we'll try an ACE inhibitor, we'll try whatever, and see if it works, and then try it again. And so even though there are a lot of medicines, there aren't necessarily a lot of objective ways of choosing which medicines to use for an individual patient, and that's where some of the promise around not only pharmacogenomics, but pharmacoproteomics and pharmacowateveromics in terms of trying to choose people a little bit more useful way. Variation and response is also the norm. The bacterial infections, there's a lot of success. Bone disease, a lot of success, but for most diseases, we get it right approximately 50% of the time. So whether it's mental health disorders or cancer or some other of your favorite illnesses, often the first therapy will work in around half of the people. And then the other half need a second therapy, either sequentially or simultaneously, or a third therapy or a fourth therapy. Gradually, there might be some benefit brought to the patient, but often it'll take a few tries to get there. And so this variation is not only a waste of resources and a waste of opportunity, because for many diseases, the first opportunity is the best opportunity. But it also decreases the trust in the whole health system as things go on. When if you can't get it right after the fourth time, can you blame someone for trying an unproven alternative approach? Because certainly the proven approach is not working very well for them. And so you can see the need to really get it right the first time from an economic standpoint, from a patient care standpoint, and from a health system process standpoint. Toxicity's also remain relatively unpredicted. I mean, of course, you tell the patient the most common toxicities, but any view that have watched TV have seen a TV ad where the last seven seconds of the ad was someone talking extremely fast, saying a bunch of toxicities that could happen to you or your loved one. And certainly toxicities can happen in a very vague sort of way. But to an individual patient, we often cannot predict who's going to have trouble and who's not. And toxicity matters a lot when it comes to the benefits of medicine. So I think most of you would agree that statins, the anti-cholesterol class of medications have been shown to have amazing public health impact. Some have argued that it has the most important health, public health impact of any medicines that have ever been developed. But that public health impact only is true for those people who take the medicine. Those people who don't take it, even when prescribed, do not get that public health benefit. It's not just a benefit by association. It's benefit by actually taking the pills. And so what we've found, and what others have found as well, is that by the end of the first year after being prescribed statins, only about one in three patients are taking the medicines as prescribed. And the reason why most have stopped, some have stopped because they don't really visualize themselves having the disease and just can't care. Some people stopped because the drugs are too expensive. But the majority stopped because of the muscle pains that are getting not, not wrapped in myelosis. Their muscles aren't shredding, but rather just the inconvenient pain that occurs in day-to-day life. And they think, oh, I just can't be bothered. I'm going to take a weekend off because I have the family reunion or I'm going to take a week off for this cruise or take a little bit of time. And before you know it, they're not taking it at all or taking it very little. And so toxicities matter not only because of the acute event, but also because the whole public health benefit of giving a medicine in the first place does not, does, is not realized when toxicities are occurring. And so there's not only an individual patient care element of this, but really a, a, a health system and societal aspect to, to toxicities and getting it right the first time. Toxicities also are, are something that happened to the patient and not the prescriber. And certainly in the areas that I work, toxicities can be extreme to the point where we don't even care about them. And of course we care about them, but we don't really acknowledge them. So with chemotherapy, one of the most common toxicities is chemotherapy-induced diarrhea. Not a topic that would necessarily talk about it at, in the morning, but, but it is, it's true. Now if, when I go to, to a study center to look at the, the toxicities, toxicities are graded from zero, meaning it didn't happen at all, to five, meaning the patient actually died from the toxicity. And so when I go to the data center, I'll ask the statisticians, just, just give me the grade three, grade four toxicities. Thankfully there's not a lot of grade five. I don't, don't even bother with the small stuff. Well I can tell you if I had grade one diarrhea from chemotherapy right now, I'd be talking to you from somewhere out in the hallway, hopefully by audio and not video. It's, it's not a, a, a trivial thing to the patient, even though as an investigator, I don't even care about it. It's not, it's not emotive enough to really beg my attention. And so toxicities are something that really have not had the full service that some of the disease aspects have in terms of genomics and in terms of other aspects of trying to figure out what's happening. You know, we, we'll sequence tumors to try to figure out which drug to give, but we won't sequence necessarily the person's germline with the purpose of, of choosing which drug based on toxicity. And I can tell you on oncology and in most areas, therapeutic selection is a tiebreaker exercise. You have two equal therapies and you're trying to just break the tie. It's not awesome therapy versus sucky therapy and you have to have a really good reason not to give awesome. It's, it's two equals often not so awesome and you're trying to decide well which of these do I give the patient and it's just a feather will shift the scale. It's not something that needs necessary amazing data and toxicity is usually that feather that will, will cause a shift to one therapy versus another for many disease areas. Now the other element that none of us want to talk about is the cost of healthcare, the cost of medications. And it's really, it's something that as, as, as academics we want to focus on, you know, shrinking tumors and avoiding Steven Johnson syndrome or some of severe toxicity. We, we don't really think about the cost element, but yet it's, it's profound for the, for the patient and often causes them to make decisions that we just cannot understand. Why would you not want to get this therapy? Well, the fact is even the well-insured have significant out-of-pocket expenses. You know, many of the, the therapies for cancer, the new biologics, the new kinesi inhibitors will be somewhere around 10 to $20,000 per month with a 10% with a well-insured person will have a 10% co-pay and it is capped at some point but most people don't have an extra thousand or two sitting around there that they wondered what to do with. And, and often people will be faced with the decision, do I mortgage my house to, to make sure I can pay for my care. So the, the economics have to be part of the decision as we go forward and, and we can't be analyzing just, you know, what's the PET scan look like for the tumor, but rather how do we put this all together and we'll come back to that point towards the end because we, we, we don't, I'm not intending that someone who's an amazing genomic scientist or an amazing clinician or amazing biochemist should suddenly become a health economic, economics person. But rather interacting with those folks to ask smart questions with sometimes with financial endpoints is, is an element that we need to be focusing on a little bit more often than we are now. Now this, this slide is really hard to see from here so you probably can't see it at all from, from there. It's from a science translational medicine article that Jeff Ginsburg and, and Jeanette McCarthy and myself put out late last year. And really the reason for showing it is not to go over in detail but rather to depict there's a lot of different areas that have to be taken into account as we try to optimize therapy based on, on genomics. There's, there's new diagnostics that are coming through and so lung cancer is not lung cancer anymore. It's one of many different elements, subtypes of, of, of cancer. The, the early diagnosis aspect is not only happening in terms of childhood maladies like you heard about last week, but, but also is happening in terms of, of predicting which diseases one might have or early prediction of one, whether one might have a recurrence or not of their disease or subsequent resistance. Is there a subclone, a resistant subclone that may be only a half a percent of the, of the total population currently that could emerge and be, taking over, over time. You know those types of things are really coming forward. So it's no, it's no longer one snip or one, one genotype equals a therapeutic decision, but rather a constellation of information that's helping inform how a patient is managed, not just at this visit but longitudinally over the, over the course of their care. And so we, we have to be thinking about a lot of different aspects and we'll hit on some of those over the, the next few minutes. Now in terms of pharmacogenomics, there is a lot of different activity happening under that name. And, and really the interaction between drugs and the genome does offer a lot of different opportunities. So there's still a lot of discovery to be, to be made. With all the, the variants that have been discovered you think there won't be too many left. And yes there are rare variants but there's also variants that are unique to populations that have not been very well studied. So, you know, for example, in some of the work that Moffat's doing, Moffat Cancer Center is doing in Puerto Rico, have found that the lung cancers, there's some variants in lung cancer that affect therapeutic decisions that occur in around 10 percent of the, the U.S.-based population, both the whites and, and African-American population, but occur somewhere around 30 percent of the Puerto Rican population. And so Puerto Ricans were not a population that had been well studied in the previous cancer studies. Now we're finding interesting back. And that's just a small little anecdote. There, but there are many throughout the literature and throughout the, the current scientific exploration where populations do have some unique features where discovery is still relevant at the sequence level. Even for common things, they're common for that population, not necessarily common in all, in all people. There, there's still a lot of difference in, in phenotypes. So whether this is the incidence of, of drowsiness after a certain medication or whether the, the incidence of a blood level or whatever it might be, still a lot of explanation going on in terms of, of, of genetic exploration. There, there's still, of course, those rare individuals. We, we have people who we give one dose of oxaloplatin and their nerves just are, are broken up based on, just on a single exposure to these drugs. There are other people who get a, a single dose or a short curse of, of carbamazepine and get Stevens-Johnson syndrome. These very extreme events are, are still a very rich source of information and, and have been the, the, the subject of not only a lot of high-profile publications, but a lot of data that has resulted in, in changes in FDA package inserts and changes in routine care in, in many different countries. Clinical trial inclusion exclusion, exclusion is now very much full of genetic information. It used to be there was a traditional phase one, phase two, phase three type of drug development and we still pretend that that's the case. But the reality is often companies are, are trying to stack the odds in their favor early on, not from a marketing standpoint, but from a drug development economic standpoint. So if you know that the extensive metabolizers have a better theoretical chance of outcome or practical chance of outcome compared to the poor metabolizers, you might do a study just in this group, compare the results to, to a competitor medicine and see whether this select group clears the bar in terms of having superior outcomes. If it does, then you can go to a traditional phase one, phase two type of model, maybe using all patients, not even selecting, but you've done that initial experiment and people who are enriched for success. And if it doesn't work there, you kill that, that drug right now, where you've only wasted a million dollars or a couple of million dollars and not a hundred million dollars. And so we see that a lot now in early drug development for all different classes, going into a genetically defined population as a almost a phase zero or phase 0.5 type of study define whether there's some proof of principle, proof of concept and then expand out into the, the formats that are required for FDA approval. And then we'll hit on some of the practice elements there, but a lot of different aspects happening now for, for this topic. The other thing is that there is more than one genome that's relevant to the patient. I'm showing you here a picture that depicts alterations in a tumor and alterations in a normal tissue, both of relevance to a cancer patient, but this could easily be an HIV patient or a hepatitis patient where the virus genome is as or more important to the treatment than the, what's going on in the patient's normal genome. And certainly the hepatitis story has taught us a lot that it's both the viral genome and the patient's germline genome that has a, it has clinical relevance in terms of some of the therapies that are, that are using. And so we cannot be thinking this, this simple religious-based approach that has been taken in the past where either you're a believer in the somatic genome or a believer in the germline and you will defend your genome to the hilt. But rather we need to be thinking about how do we improve care for this patient who happens to have both within them and both need to be accounted for. And so this, this approach where we're taking into account both genomes simultaneously is really being a much more rewarding strategy. Risk benefits not just benefit or risk in, in isolation. And so that's an important element as, as we go forward. Now the way pharmacogenetics or pharmacogenomics is being applied is really rather simple at this point in, in routine practice. It's often used retrospectively, clinically, to explain an untoward event. So someone received 5-3-uricil or capesidobine, a 5-3-uricil-pro-drug, had a very severe reaction and you want to know was it dihydroprimidine dehydrogenase deficiency that caused them to get that extreme event. Because if it was, you now know to either not use that drug or use extremely small doses. Whereas if it wasn't that event, you, you now need to use a different strategy in terms of, of managing the patient. And so that's a common example. There are also areas where there's low utility that are, end up being requirements for insurance coverage. So if, if you have a colon cancer that has a mutation in the KRAS gene, you will not benefit from some of these expensive antibody therapies. And therefore, without evidence of genotype, insurance companies will not pay for the medicine. And so, as you can imagine, as an economic requirement, that test is done very faithfully, because there's not just clinical, but also economic reasons to make sure that happens. You have evidence for dose selection in terms of, of whether someone needs a normal standard dose or a higher dose. Therapy selection in terms of whether one gets a, a, the, the most commonly used medicine in case of, of clopidogrel for, for stent placement in the heart. Clopidogrel or plavix is a, is a commonly used medicine for there. But in patients who have mutations in CYP2C19, they can't activate it as, as faithfully. Often people will use a, a different medicine that's more expensive, but is, doesn't, but bypasses this particular activation step. And then you have preemptive examples. So this particular genotype is used for the HIV drug Abacavir, but there are other HLA markers for severe hypersensitivity reactions. Some of them occur more commonly in East Asian populations. And you'll see in countries like Taiwan and Thailand and, and China that these tests are routine in terms of the management of patients receiving carbamazepine, some of the HIV drugs all appear and all. They're paid for by the government. They're a routine test because it's such a high frequency of event, whereas in the United States it's not used as frequently, used more commonly on the West Coast than the East Coast, but, but they, these tests have not, have not become as popular in terms of clinical management. Now in, in terms of pharmacogenomics in 2014, there are quite a few examples where application is happening. It's not something that may happen someday, although it will increase in time just like every other as knowledge, as knowledge goes forward. A number of these examples are, are tumor aberrations. Some of them quite old. The, the HER2 example is, is quite old, where a tumor abnormality and copy number and sequence in rearrangement will lead to a selection of a medicine or a de-selection of a medicine in, in, in these, these cases. There are also examples that are associated with toxicity that will require altered dosing. Examples associated with hypersensitivity reactions. Examples in terms of, of drug therapy selection. Many different types of examples that are used. Now what's shown on this slide are the examples that have made it into the dosing and administration section of the FDA prescribing recommendations or the package insert, as they're more commonly called. Now the reason that's important is that there are 140 different drugs or more than 140 now that have genetic information somewhere in the FDA insert. But this list here they're in the dosing and administration section, which is the section that is supposed to be read by prescribers. It's the section that is read by the, the iPhone apps that, that you use to prescribe. It is the section that's read by the insurance companies and unfortunately it's the section that's read by the litigators. And so we see a lot of, of, of litigation emerging where someone did not do a genotype. Something bad happened and then they can use that, that result to beat up some poor sap who is doing the best they could. In some cases the, the event happened before the FDA even acted. But yet that doesn't stop the litigation attorneys in terms of, of trying to take a, a genomic driven approach for, for that. But there, there are examples now and the list increases as, as the data allows. Now I'm going to hit on a couple of different points over the next little period of time. One is around discovery because there's still a lot of discovery to be done. If you look at the FDA approved drugs, sorry, if you look at the, the top 200 prescribed drugs there's only a, a, about a fifth of them that have had serious genomic analysis of any type. At least in the public, published literature. Now many of those are old drugs and so there is no sugar daddy to pay for that study. The NIH has not been a, a big funder of, of pharmacogenomic studies. There, it's been more in, in industry or in foundations. And so many of the old drugs are kind of old and boring and, and have not necessarily received that, that type of, of evaluation. And so there's a lot to be done still. And even some of the examples where work has been done there's still elements to, to be, to, to be defined. Now I should like to show this slide as a way of reminding ourselves that we know something, but we don't know enough. This is an anti-cancer drug. It, it goes into cells and it's, it's pumped out through an active transport. It's inactivated through P450s in the liver. It's activated in the plasma to this metabolite, which is pumped out, which is inactivated, which hits a cellular target, cell death occurs. I mean, look how smart we are. I mean, we're geniuses. And if I had a better graphic artist, I'd be even smarter. Except here's the real pathway. Especially in the area of pharmacodynamics. It's, it's like, I don't know if it's Yogi Berra or Donald Rumsfeld, but we, we know what we know, but we don't know what we don't know. I mean, we, we have a situation where someone has looked at these particular genes and has seen some sort of effect in cells, in mice, in man, somewhere. But it's not as if someone has really asked the question which genes are most important and are regulating. Let the biology tell us where, where to go. And so it's often the situation where I've got an assay for CYP3A5 running in my lab. So I don't care what the question is. The answer is CYP3A5. And, and we see a lot of that even in modern science, where, where people, you know, I've invested in this assay. So I'm going to run this sucker. And, and, and we need to be, be stepping back from that. Because often it's leading us down the blind alleys. So a lot of discovery is still going on in, in mouse, in man, in family studies, in, in all sorts of different approaches to try to, to try to help. But really a not a lot has, has been done. We're very early in this field. The term pharmacogenetics was coined over 50 years ago. But the, the science in terms of really trying to aggressively define which genes are important is, is rather new. As of yesterday, there were 2,228 genome-wide association studies in the NHGRI GWAS catalog. 73 of those studies had a drug-related phenotype of some sort. So less than, only 3 percent or so. Very few of them had a large sample size. A minority of them found no significant hits at all. There's a, just, just around half had a replication cohort of some sort. And that's an improvement because a few years ago when I looked, it was a lot fewer that had replication cohorts. But even though you have this, this mess and this, you know, we've hardly ever even started to try, there have been 11 of these studies that have contributed to package insert changes at the FDA. And so there, there are some, some, some bits of gold to be pulled out of the mind. But it, it takes effort to, to get that out. And, and really we're just starting to try in terms of finding what are the genes that are important. In some cases there will be no genes that are important. Either because the, the effect of any one gene is so small, or because it's post genetic effects that are, that are critical. Or there'll be some cases where it gets all the way to the point where we're, we're driving patient care based on, on some of these changes. But we, we have to do those studies. So there's a lot of work still to be done. We've also stepped back in some areas and, and tried to do things in a little bit different way. So if you can imagine where have the big successes happened in terms of gene finding, mice have been a huge success for, for disease in general. Family studies have also been a huge success even in the next gen sequencing era. Family studies have been a very valuable source of finding real genes that hold up as being clinically important. And, and so that's great except family studies are tough to do with certain therapeutic classes of medicines. Like anti-cancer drugs in general are, are very hard to give to normal volunteers. The, the risks are just too great. And you're bringing volunteers into a, into your, your lab or your clinic to do a study, you know, sorry about grandma's, not something you want to be saying to your volunteers because you gave them nutripenic fever and died of sepsis. You know, it's, it's not, it's not like there's an inconvenient rash. It's something a little more than that that often will happen in these cases. And so one of the things that, that we and others have done is, is tried to step back and say, well what else can we do? And one of us is using immortalized B cells, EBV trans, transform B cells from large families. Now some of these families are these, these CEPF families, the name shown, shown there for you French speakers, that the, the benefit of those is first of all, they're multi-generational and have large numbers of children. But, but also much of the human genome project has used these cells and therefore the genomics is already in place and yet, and so you just have to do the phenotyping and you get the genomics for free. And now there are resources through the NIH and otherwise where large numbers of unrelated individuals with genomic data available and intact can, can be brought into the same scenario. And so one of the things that you can do and I'm showing you a 96 well plate because it's, it's prettier, we do mainly 384 or 1536 well plates. But one can, can do two different drugs on a plate in quadruplicate with increasing drug concentration, several different types of controls on there. And in case of cytotoxicity, do a 72-hour assay with a, a, an oxidative stress dye as the, as the phenotype. And you can see that some cell lines can have a very rapid killing. And these are, are three separate experiments, not three separate plates, but three separate days or three separate occasions with the error bars shown there. So it's, it's pretty tight replication. Admittedly using robots and bar coding and things that allow you to get better, better data like that. So in this particular cell line, very rapid killing with increasing drug concentration for this cytotoxic drug. And in this cell line, same drug, same concentrations, very little killing to the point where it never even reached 50 percent killing rate for this drug. And so you see this type of variation. And then one can ask some really fundamental questions. One of the questions is, is the trait heritable? Now it seems pretty stupid that we didn't do this before. It's embarrassing to think how many millions of dollars I personally spent on genomic analysis without asking the question, is the trait actually inherited? Now heritability is, is a, a phenotype that, that can be influenced by a number of things. It's not just a, a useful predictor of whose, whether there's a gene involved. For example, I can't really see because of the lighting, but I imagine that the majority of you have two arms. And there were genes that were involved in, in that, even though the heritability would be quite low. So you can see that heritability is, is a measure of, of variability as well as, as genomic influence. And so one can ask that question using the families, the families of cell lines. And this is data from, from 14 different families, about 150 different participants looking at the 29 most commonly prescribed anti-cancer drugs. And I wish it was 30 because I hated it as 29, but there was one of the drugs where he had solubility problems, we couldn't trust the data. So I, I hate that it's not 30. But what one can see without even being able to read down here is that there are some drugs with very high heritability up in the 60% range. Some drugs very low heritability similar to the, the controls, the vehicle controls. And so this, this gradient is present. And one can now say, well, not necessarily that these would not have any genes involved, but certainly you can start prioritizing up on this end of the, of the scenario. Now at the top of the list here is a drug called timazolamide, which is used for brain tumor treatment. It's an alkylator agent. And so we, we then did a, used a collection of, of 563 unrelated individuals, took their cell lines, looked at timazolamide in, in, in that environment, did a genome wide association study using in vitro data. And what we found is a hit here on chromosome 10. And even without the green lines you can probably see this something we, we, we, that just came out a, a, a couple months ago with Chad Brown as the first author. And you can see here this, this hit. Now the good news is that this, this hit was methyl guanine, methyl transphrase. A gene which repairs DNA addicts. Perfect. The bad news is that biochemists had already shown this gene to be involved 10 years ago or more using traditional biochemical analysis. So the positive spin of course is that we validate our approach by finding truth. But the reality is in this case we, we found something that somebody had already found before. But we now can take this and look at large numbers of, of, of other drugs where we have hits that have not been associated previously with these drugs as the start of a series of, of biochemical analyses using SHR and A, etc. To try to credential which of these genes that we, here too for had not included on our list of, of important genes really have some impact in, in terms of the effect of these drugs. So this sort of discovery approach is just one of many. But the idea that we still need to do discovery is so important because you know often we think oh well you know the genome era is well, we're well into the, now the second decade. All the discovery has been done we just need to now apply it. And that is certainly not true. We do need to apply it. But there's a lot of discovery still to, to be made and matter of fact there are so few people trying to do the discovery that it's no, it's no surprise that we're so slow in, in terms of advancing the science. A second aspect is validation and we really need robust data sets. And, and what, what we've found is that there are very few high quality biobanks out there. There are biobanks in terms of flesh that are stuck in a freezer. But in terms of high quality annotated data there, there's really very little. And, and it's shocking how little is happening within the NIH clinical trials portfolio. There, there are some areas like cancer where now the NCI is funding either blood or when possible tumor accrual. But many of the other institutes have not mandated a, a collection of blood and other, otherwise. NIH has done a good job but there's, there's still a lot of work to be done. And so there's a lot of missed opportunities. One of the things that we did a few years ago in cancer area is start integrating blood sampling when possible tumor sampling. And you, you have scenarios where instead of 46 breast cancers from Tampa, Florida, you'll have 4,600 samples from centers all across the United States and Canada where you have captured the variability of multi-center treatment but in the context of a prospective clinical trial with audited data for both toxicity and efficacy, sensor review of the imaging, all those kind of quality control measures one needs to trust your phenotype in terms of, of analysis. In, in terms of some of the drugs there you have some very nice grading with well-defined criteria that have been put forward by the NCI or by the other institutes. So you can really have uniform measurement of toxicity in, in there. So here's a study. The clinical trial was published two years ago. Dan Hertz has a paper on neuropathy that is, is coming out. There's more data to be published soon. But this was a study in prostate cancer where dosataxyl, a chemotherapy drug and placebo was compared with dosataxyl and bevacizumab which is an anti-vegev, anti-vascular agent. And the, the bottom line clinically was that these two arms were not different in terms of survival. There was a difference in terms of disease progression but, but not in terms of survival. And so one can go in and do genetic analysis. Here's the most common toxicities. You got neutropenia, you got neuropathy, hypertension, thrombosis, hemorrhage, et cetera. And you can use those phenotypes to try to analyze things, things further. I'm skipping some of these in the interest of time. And so one can see a study design where one could now do validation or even discovery or look at patients who were treated on the trial who experienced neuropathy or who did not experience neuropathy in, in terms of that particular phenotype. Except it's not as simple as that. There are many other things that need to be taken into account. Competing events as they're called. And so it could be that the patient's disease progressed prior to having the chance to get neuropathy or that they died or they had some other toxicity or withdrawed withdrew from the study. There are other events. And so the statistical modeling needs to catch up with the clinical reality. We have decent models for a traditional type of strategy. But the competing risk analysis models that are out there are okay, but not nearly what we need in terms of trying to definitively answer these, these questions. Now I'm going to skip over some of this in the interest of time, but the bottom line is that on this trial patients got neuropathy shown in the red line or this bottom line for those of you that are colorblind. But you had other toxicities like death and progression or other adverse events that occurred much more commonly. And so you have to take into account these competing measures. You can't just rely on a yes, no type of phenotype in terms of that. We also often need to take into account dose because many of the toxicities are dose related in addition to in addition to just the presence of the drug. And so the level of sophistication that needs to be put in is the same as disease genomics. You know, diabetes, no diabetes is really not very useful. Diabetes early onset or timing to onset or diabetes well controlled but still having kidney damage. There are other phenotypes need to be brought into play to really define the patients. And the same is true with drug effects. We need to be looking much more sophisticated in terms of how we're defining the drug effects. Not just saying a yes, no and then wondering why we don't find anything. Now in terms of this particular analysis this is a most of you are familiar with genome wide association analysis has been talked about over the last few weeks of this course. This is going along is chromosome one, chromosome two, chromosome three, etc. And on this axis is the negative log P value. So the higher the value the more significant each one of these dots is a region of the genome where there was a single nucleotide polymorphism that gave some level of data. And these are so called Manhattan plots because when you have a positive finding you have some sort of gleaming spires like you'd see in Manhattan, New York. Unfortunately often it looks more like Manhattan, Kansas. And in this particular case I had to put red circles around the dots or you wouldn't even see them. And yes they were above a certain threshold statistically but we're often with these phenotypes they are complex traits. They are not a simple Mendelian style trait. And so one has a number of genes contributing a little in terms of its prediction. Now when you look at the list of genes that are there you see that some of them are when adjusted are genome-wide significant. Others are almost. But every one of the genes has that perfect story for why it should be included. And so it here's stabilizes something in Chocomery tooth which is a peripheral inherited peripheral neuropathy syndrome. Perfect. Peripheral neuropathy peripheral neuropathy syndrome. Beautiful. This gene here is involved in the dorsal root ganglia and maintaining neuron. Oh perfect. Neural outgrowth. Oh it even says it in the name. Great. So what we find is that our statisticians will label these gene one, gene two, gene three for us. And by doing that we have a much more objective discussion about which is important and which is not. Remember Pat Brown talking about the early days of expression arrays where they did gene expression for breast tumor and normal breast ducts and they got the list of genes and they went spent the afternoon going through each gene and talking about why this one made sense and that one made sense. And then the next morning the statisticians came in and said sorry there was a coding error here's the list of genes. Every gene had made perfect sense before but it just was not true. And unfortunately the way we've named genes in the genome anything that has cell death in the name or something like that makes a perfect story for any phenotype you care about. And so yes we find genes yes they seem to have some biologic plausibility but there's still a lot of work to be done in terms of replication and validation and then implementation in terms of how we use it in practice. Skip over that in the interest of time. Now the other thing that one can do is not only take advantage of the clinical trials but take advantage of the health systems. So this is an example of something that started at Moffitt by Bill Dalton back in 2006. I had nothing to do with it at that time so it took it no credit for it but what Bill did is he developed something called total cancer care which it's really the most of it is total cancer collection. There is a care aspect to it but what's happened is that you can go in and from day one patients are enrolled in terms of clinical follow up clinical data warehouse retrieval but also in terms of biobanking. And so for example as of a couple weeks ago there were 105,000 tumors that had been banked at Moffitt all with extensive clinical information available longitudinally. Consented from day one to allow a whole genome analysis if you want to do that in terms of the way forward and there have been some of these for example 16,000 of these have had gene expression analysis done 4,000 have had targeted exome there's been a small number that have had whole genome etc. So one can if one builds a biobank longitudinally and lets it grow with purposeful investment one can start reaping rewards. Now most biobanks were designed for deposit and not withdraw. If you put your money into a bank that was for deposit and not withdraw you'd be pissed. You know you want an ATM on every corner you don't just want to be able to go in and get your money out you want to have it convenient. Matter of fact you want an app where you can just do the transfer on the app you don't even want to have to ever go to the bank ever again. And biobanks traditionally have not been designed with people in mind. They've been designed with sticking pieces of flesh in a freezer and we need to be rethinking how we're doing biobanking because biobanks are great if you want to know tumor versus normal or one tissue versus another tissue. But if you want to discover what's associated with some sort of outcome toxicity efficacy whatever you need to have high quality phenotype. Phenotype rules the day. You can sequence whatever you want but if you have crappy phenotype you'll learn nothing. And so we really need a lot more attention placed on that in order to improve the infrastructure that we have in this country for doing this sort of work. Let's get that in the interest of time. Now the last part I want to spend a little time on is application. And some of this will be spent on traditional application. What are we doing for the patient that is in clinic today? And then some of it will be application in terms of public health use of pharmacogenomic data which is a little bit different than maybe you were expecting but also an opportunity in terms of trying to help developing countries in particular make decisions about which medicines are available in their countries. Now I showed this list before lots of different types of drugs that are available. And what's interesting is that many of these drugs have very little application data that's out there. There are very few implementation science studies that have been done with any of these examples. There's been some quality of data that has got into a prestigious journal that has led the FDA to make the change but traditionally not a lot of how do we actually implement it in routine practice. Now one example that's a bit controversial and I'll tell you about that but has really taught us a lot is one that I'm going to show you here. We've published some on this and then there's some of this data is not yet published but the concepts were important enough that I thought I would take that risk. Now tomoxibin is a drug used for breast cancer. It's an old drug. The drug itself is not so potent in terms of an anti-estrogen matter of fact it's not really potent at all but it needs to be activated to metabolites which are potent anti-estrogens. And when I trained for hydroxy tomoxibin was the main drug that active metabolite and there was a bunch of different enzymes that were involved in activating it and so there was not really a lot of variability and therefore that was the end of the story but a few years ago almost a decade ago now Barrett Stearns and David Flockhart when they were at Georgetown saw a woman who was had breast cancer was receiving tomoxibin was getting the hot flashes that one gets with the perimenopausal syndrome as you block estrogen you get this type of syndrome many of the patients have it. She also had clinical depression was seen by I'm not sure if it's a psychiatrist or a family medicine person but anyway was given an anti-depressant and her hot flashes went away and then went away very quickly. So if it was me I would be quite excited. Hot flashes went away somewhere around the next four or six weeks the depression will probably benefit anybody with Scottish blood loves two for the price of one. So fantastic right? Well Barrett's a KG Israeli she didn't think that was right that something that took four to six weeks to work for depression would right away affect hot flashes something was going on and through a number of years of very difficult analysis they identified that there was a metabolite that they called endoxifen it was a known metabolite but not a prioritized metabolite was a very potent anti-estrogen it was also something that was formed mainly through this two-step process down here and the reason it was relevant is that the anti-depressant that was given blocked CYP2D6 so the reason the hot flashes went away is that active drug was not being formed anymore. Now the good news is you block formation of active metabolites you don't have hot flashes. The bad news is you have no anti-cancer protection and so what we saw dramatically is as this was presented in June of a few years back at the American Society of Clinical Oncology Meeting almost immediately looking back at prescribing data almost immediately people stopped using that those types of anti-depressants in patients receiving tamoxifen it was just a wholesale they stopped doing it patients that were already on them were switched to other anti-depressants they still used anti-depressants to try to fight hot flashes but not the ones that block this step of activation and so it was very widely accepted to this day, so Liz, that you do not want to mess with CYP2D6 in terms of tamoxifen therapy now the funny thing is kind of a genetic exceptionalism type of story so with a drug interaction it was blindly accepted that you don't want to mess here and therefore we will not use those drugs but yet 10% of you in the room and 10% of you watching are missing this gene either missing both copies of the gene from deletion or having a very low or no function based on point mutation and so those folks you think will be as or more risk of a bad effect than the people with drug interactions because drug interactions might occur and it's very much blood level dependent etc whereas genetics it does occur it's there's not a lot of variation there and so there have been a number of studies that have shown a story similar to this one where the people with two copies the extensive metabolizers have the better outcome compared to people with one working copy or no working copies of this particular gene so the poor metabolizers as they're called have a worse outcome now while this is an independent effect compared to other factors you'll see that even the good risk group every time there's a little blip here someone's breast cancer has recurred and so it still does happen it's not like this is the gene that cures breast cancer but it does seem to have some effect and there have been a number of studies showing that this type of scenario there have also been some studies that have been that have shown that this does not occur now in some of the cases a very high dose of Tvoxoma was given or a lot of extra chemotherapy was given in some cases the tissue of in one case it was a prevention study so a different scenario in terms of incidence in the last two cases tumor DNA was used and so there's some confusion there about the relevance because about 30% of breast cancers have a deletion in the region of this gene and therefore when you genotype with breast cancer tissue what are you really genotyping and there's still a lot of controversy about whether these studies told us that this gene is irrelevant or whether the studies showed us that we need to use the right tissue so still to be defined with that but this idea that we have a group that does well a group that does poorly they can stay on the one pill a day 20 milligram Tamoxifen and they need to have something else easy except 40% of women are in this group here this middle line and so as oncologists started to do testing many of us started to get phone calls saying hey what do I do with these folks and the reality is we didn't know we didn't know what to do and so some studies that were done and one of the papers came out a couple years ago there's another paper that is is under review right now on this the first 119 patients have been published and as I mentioned as I showed you down here additional 500 patients is now being reviewed what we did in this study was really simple we took patients that have been on Tamoxifen for at least four months so based on the pharmacokinetics they reached steady state we then measured active metabolite levels shown on this axis we also did clear level genotyping clinical grade genotyping for CIP 2D6 and what we found is what others had found there was a statistical difference between the intermediate metabolizers and the poor and the extensive metabolizers people that have two normal copies or only one working copy of the gene has statistically different active metabolite levels no surprise but it was nice still to see that what we did is these folks here they stayed on one pill a day and four months later there was a slight decrease not statistically different probably due to adherence in terms of taking the meds but not a lot of change that occurred there these folks here with the low levels we did something really simple we had them take two pills instead of one nothing earth shattering nothing crazy just that and the FDA approved dosing is between 10 and 40 milligrams so they went from 20 milligrams a day to 40 milligrams a day we didn't even have to file an IND because it was within the approved dosing and what we found is that there was no longer a statistical difference between blood levels at that point we had normalized blood level now two additional studies one from Tokyo and one from New York City have now replicated this finding so it's something that definitely does occur there's a significant difference here and we can normalize it based on genotype derived dosing one pill versus two really simple stuff what's funny about it is that this sort of data is exactly what's behind most of the FDA dosing recommendations so if you dose based on kidney function it's a pharmacokinetic study that derived that dosing pharmacokinetics were 50% different therefore you give a 50% different dose drug interactions are almost all pharmacokinetic based interactions and the goal is to normalize by taking into account drug interactions organ dysfunction age whatever the factor might be and so from that standpoint we've now defined and others have replicated a way of normalizing blood levels what's controversial is whether this this sort of normalization will impact survival at all and so we went and enrolled a total of 500 women from across North Carolina I was up in University of North Carolina Chapel Hill at that time we enrolled 500 women the outcome now data is now maturing we need five-year survival data so it's going to take a few years to do that but we have been able to show that in the 500 patients we indeed can replicate this this finding where there's a significant difference at the start that one can normalize based on this interaction the other thing is that we were able to do this study in 64 of the 100 North Carolina counties we took this study out of the academic center out into the community and we're able to enroll a genotype guided therapy study out there just fine matter of fact we thought we would enroll 100 patients in five years we enrolled 119 in four months expanded it to 500 and enrolled 500 in 14 months and the reason why is it was a very simple study it was a simple concept that any oncologist could understand and any patient can understand and we had the weird scenario where we didn't have to advertise the patients did it the very first patient that came in I've met with her along with Billy Irvin and some of the others that were involved told her about the study and she's like yeah I want to be involved with that sure fine she left the room we then called in the next patient and we were in quiet phase just making sure we had IRB approval but making sure the forms were correct whatever the patient came in and said I want to be on the DNA study like how did you find out about it oh a woman came out and went around and told everybody in the waiting room patients think we already do personalized medicine they think we already do this stuff they want to have their care to be shaped after their body the way that they their liver acts or whatever it might be and so patients are on board with this sort of stuff but we need to do the trials to show that it really is the right thing to do and what we'll find out in terms of that another area that we've that is important is with opportunistic infection so whether it's HIV or or or cancer or whatever it might be often it's not the the disease that kills the patient it's opportunistic infection and certainly in the case of myeloid malignancies acute myeloid leukemia in particular you have a a very high risk of fungal infection and many patients will die of fungal infection and so there have been studies that have shown that sorry for the complex drawing here that the vericonazole an antifungal drug needs to be works by itself but it can be inactivated by a number of genes including CYP2C19 that show you here and we already know that people who have two bad copies of the gene have very low blood levels and then there's some people that have extra copies of the gene that have very high levels and so those people have high levels and often get hallucination and other effect that there's others that that can chew the drug up really fast and I and get they do not reach therapeutic range in terms of of their blood levels and indeed this is an ugly slide that I need to and apologies to the authors but this is an ugly slide that they what we found is that it normally though the rate of ultra rapid metabolizers is is somewhere around 20% or so and yet 80% of the patients with with with unachievable that did not achieve therapeutic blood levels had this particular genotype so there's an enrichment the people who chew up the drug too fast can't get therapeutic levels and are the ones that have fungal infection now the reason for belaboring the story is is we we wanted to go and start implementing this in in our cancer patients in our leukemia patients in particular but we felt you know we need to do an economic analysis to see whether this is is really relevant what we hadn't realized is using our own institutional data a person who gets a fungal infection costs an extra $30,000 well $29,500 and compared to someone who does not get a fungal infection and so we did a number of different economic analysis Neil Mason from our group did this we could not find a scenario where it wasn't cost-effective to genotype everybody because if you if you prevent one case you've paid for everyone in spades and so often we have to include these economic analyses in addition to the science to make sure that we can make the case for that you know if I could keep a few leukemia patients from getting a fungal infection but bankrupt my entire institution I'm not doing anyone any favors and so we have to bring in not only a science of implementation and the science of discovery and validation but also get the economists and others involved in terms of the application now I also want to mention a little bit about the complexities that are becoming normal in terms of of cancer care you know it wasn't that long ago that we would say oh there's a tumor in the bowel therefore this is bowel cancer or maybe we get fancy and call it colorectal cancer that's that's great well then you can look at it under a microscope and stain it and say oh it's an adenocarcinoma there's some ducts there so there's a glands there's a so it's an adenocarcinoma of the colon very fancy very fancy oh well now we can genotype it in this case for k-rass gene and say oh it's a k-rass mutant it's a code on 12 k-rass mutant adenocarcinoma of the colon wow we're starting to get pretty fancy here that's right well now we start getting into some of these realities and this is a I'd use my iPhone to do take a picture of this report a sequencing of one of our patient's tumors and maybe you can see up here that there is an abnormality in p53 in ep300 and ddx3x gene is lost well that's great and now we have to go and figure out well what are those genes and and should it matter and the report from the company says that this really has no relevance to those of FDA approved therapy and there are no clinical trials involved so are we really any smarter and there's a list of a bunch of other genes that are abnormal then no one has a clue what to do it because this these variations are just and the handwriting's on here because we get together and try to figure out what in the heck do we do with this stuff so it's no longer the case where it's a simple little colon cancer or a simple little leukemia we're getting a high layer of complexity such that the informaticis and the biochemists are heavily involved in trying to help us understand what to do with the clinical data folks that were used to be in another building and we never met except in the cafeteria are now an integral part in terms of how we manage our patients because we need their expertise in terms of of trying to interpret it's really we're at the point or the early stage of consensus opinion not at the stage where we have definitive data saying yes if you have a HIST1H1E mutation at codon47 we know exactly what to do we don't have a clue what to do and we need to dig into it and find out are there trials is it likely to be important what do we actually do with all this stuff another switching gears from the tumor side another part is what do we do with the rest of the world we that's great that we can do genotype guided whatever for cancer or for HIV or for whatever your favorite disease is but what do we do in most of the world most people don't have access to the genome most people don't have access to any of this stuff we've been talking about in this whole series so what does it mean and modern therapy has been a key component of improving health and really is a sizable part of most health budgets in the developing world most of the time the buildings are not super expensive the people are not super expensive the cost of equipment is minimal don't have a lot of the super expensive equipment but the medicines even the cheap generic medicines from from indio or other parts are still expensive in terms of the proportion of the health care budget and when you look at selecting medicines often it's a combination of clinical consensus access to and cost and familiarity and so you have these sorts of scenarios and medicine prioritization is really a high stakes undertaking in most of the of the developing world when you look at the WHO essential medicines list which is the national formulary for the most countries use outside of the richer countries what you see on there is that there might be five medicines for your favorite disease and if you can only afford one or two how do you pick the data up until even 2014 is is heavily skewed towards western europe united states and australia we have data that's coming from certain populations of the world but they don't necessarily benefit or give direct data for for the rest of the world and so instead of having data for the individual patient we have data from certain parts of the world that are inferred across the rest of the world so if if you're in this case Nicaragua or in other other parts and you're trying to make a decision about what is your what medicines you pay for you can't pay for them all you can't afford that so how do you select and Nicaragua is not a participant in the phase three clinical trial networks that are currently out there there have been no trials no patients enrolled from Nicaragua in any of the FDA approved trials over the last 15 years you have nothing to go on in terms of trying to make your decision at the the ministry of health level and and so what we've been doing is it others have been doing is trying to look at can we use pharmacogenetic data to augment a decision that a that a country's ministry of health or health authority might be making to provide some local context for that and and the way we've been doing that and we've worked in 104 countries so far is to identify the most common groups within the country that might be ethnic that might be racial that might be religious whatever the the country decides are those groups enroll volunteers from these different groups and and look at genes of gene variants that have been shown to influence toxicity or efficacy for medicines that are on the WHO essential medicines list not alterations in pharmacokinetics but alterations in actual dosing toxicity or efficacy and those alterations have had to be found at least two separate populations in order to be included on the the list this is something from the pharmacogenetics for every nation initiative or P. Genie as it's as it's called think P. Diddy except P. Genie and so one can go in and if you take an example with macaptor purine which is used for arthritis for inflammatory bowel disease it's also used for childhood leukemia but that's a the smallest therapeutic category happens to be the only after you approve category but it's the smallest therapeutic category in terms of its use there are are three different genetic variants in the phallopurin methyltransferase gene that inactivates that influence the inactivation so if you have one of these different alleles you can't break the drug down very well you get extreme toxicity and need to be hospitalized or at the least come off the medicine and it's been well shown by multiple groups now that there's very different dosing depending on whether you have two normal copies of the gene or two abnormal copies of the gene it's almost one-tenth of the normal dose that one would take in that in that scenario so that's great so this is some some of the initial data from a while back now and if you see green that means the data is similar to what's seen in the U.S. white population and the reason why the U.S. whites for the comparison is not because I'm from the U.S. and white but rather because the dosing and safety data was almost exclusively done in the initial stages in U.S. populations phase two and phase three then go out to other parts of the world and so you can see there's many countries that have green some countries have light blue that means the genetic risk is one-half or less of that seen in in the U.S. white population and then there are some countries Bulgaria, Ghana and Peru in this particular slide that have more than double the genetic risk for this incidence and so the connection between those three countries is a bit perilous there's Cocoa and two of the three and who knows what else but those are three countries that really stood out now and here is in a continuous variable if one looks within Ghana in West Africa looking at some of the more common sub-populations within the country what you find is that all of them have about a 10% incidence of the severe risk very high incidence and often what people say is well geographically why don't you just look somewhere in Africa and then you'll get African data and somewhere in South America you get South American data well here's the data from one of the Nigerian populations the frequency of genetic risk for this particular example in this population is half of that scene in their neighbors or two small countries in between them in Ghana matter of fact the Nigerians in this case the risk was more similar to the UK Caucasians and the US Caucasians than it was their neighbors in West Africa and so one can't just take a blind look just get the region and try to get it right here's another example this is CYP2C19 it activates for akanazole an antifungal drug as I showed you a couple of slides ago it's also involved in the activation of plavix or clopidogrel for cardiac disease and what you can kind of see up here this is CYP2C19 star two variants here's one Ghanaian population with a very high frequency of this variation of this mutation and here's another Ghanaian population the Fonte versus the Ewa very different frequencies even within the same country and so and then you can see some the Kenyan populations and Nigerian populations next to them so the the point being that by looking within an individual country working with the Ministry of Health trying to define what is the level of risk within the populations that they define one can start getting some data it's it's not so much a clear decision based on this but it's it's tiebreaker type data we're using it almost like a currency converter you know how many pesos equals a dollar how you know how much which drugs are more likely to be safe in Ghana versus another and using this more more broadly and one can use the data in terms of privatization of or surveillance so in the case of of a high incidence of liver toxicity from isoniazid one is still going to give isoniazid there's not a good alternative yet but you'll monitor more carefully and and that is this data we've we've some of the data for liver toxicity risk from isoniazid with countries like China I'm using this data to help define the way they monitor these patients in other cases we can have clinical algorithms for the available drugs in this case for rheumatoid arthritis where based on the genetic data within the country one can start breaking this down into a clear decision this is for a small eastern Asian country called China where one can take the drugs that are available and help prioritize them in terms of of level of risk so for example methotrexate they have a very high incidence of a resistance gene with thymolacentes and it was already known that methotrexate didn't work as well now they figured out why and so one can use this in terms of coming up with priorities for their medicines we can also look more broadly this is data from aphometrics DMET plus chip looking at 7,000 individuals across 40 different countries and you can see that here's the average predicted warfarin dose on this axis and you can see that in some cases the average dose is extremely small other cases that's very high based on the different geographic continental separations GI risk from amidiaquin a malaria drug again high risk low risk a lot of variation within a continent risk of semastatin muscle toxicity same type of thing we want to start putting together these sorts of maps and reports for the Ministry of Health to now not just take the WHO essential medicine list and say we got to pick one of these but use this data to try to say you know what we can now prioritize based on this and we can only afford the following three so we'll use it in that sort of manner the closing here the key thing is that we've become decent at discovery and validation we still have a lot of work to do in terms of health economics integration health systems you can tell this slide is old because there's a blackberry on there and for you young people blackberries used to be like an iPhone probably haven't heard of them but they're in and acid development a lot of work still to be done and I'm going to finish up with this particular story from a friend of mine in North Carolina at the time he was 44 years old he's the chief scientific officer of a a biotech company in North Carolina he was born with what he calls a frog heart he has an AV block due to this congenital heart defect needed to have a pacemaker placed but because of his anatomy he had to have his chest cracked to place it and so he told the cardiologist the CT surgeon the anesthesiologist and the meeting team of these folks here that he'd had he'd had an executive physical that he had pharmacogenetic analysis and that he couldn't activate oxycodone very well or codeine and also had to have some different dosing of warfarin that's fine they noted that and went on with their way successful surgery successful placement rather of the the pacemaker in the recovery room morphine was giving him some decent paid control four out of 10 in the scale he has chest crack so it's you know sort of a painful thing he moved to the coronary care unit he was switched to oral medicine stoxicodone and had very severe pain was basically ignored he was a wimp you know he's in pain just needs to buck up he called me from the ccu you have to come and rescue me I called one of the cardiologists at that at that university who had had trained with me when I was a wash you and said hey come on you gotta go save this guy before he got in there one of the medical students and one of the pharmacists were pre-rounding saw this this man in severe pain talked to him found out he was a poor metabolizer precip 2d6 so his ability to activate oxicodone and other pain meds is not as good as most people switched him to a different med in this case hydro morphone which gave him a much better degree of pain control still five out of 10 but he had his chest crack you know now this happened at one of the top five cardiac centers in the united states a very high profile institution that happens to be eight miles away from university north carolina rhymes with duke a phenomenal place I literally have gone there for cardiac evaluation because it's such a great place the anesthesiologist who was involved literally wrote the book on post-anesthesia pain and yet world famous cardiologists cardiothoracic surgeons anesthesiologists incredibly smart fellows residents and interns didn't recognize the data when it was in front of them now if this had happened at UNC it would have been just as bad or worse and we did a lot of cross training based on this case where jeff ginsberg came over and trained us at UNC and I went over and helped train them at duke but the bottom line was even with the data in our face we don't always recognize when it's ready for use and so we can't have a scenario where we just publish smart papers we have to be thinking about implementation and the best scenario is one where no one has to know about it because it's baked into the electronic health record and this sort of data is a hard stop to switch over to a different medicine and no one has to remember anything because the computers do that and that's something that is now in place at minicenters is now in place at duke and it is something that is applied so I'll finish with this slide you know back at the last Olympics the one in London usain bolt did not win the 4x100 relay Jamaica won the 4x100 relay men's relay usain bolt ran a phenomenal leg just a terrific leg but so did three other guys and if they hadn't ran terrific legs Jamaica would not have won it's the same type of thing you know when you go to pass the track where they're practicing they spend hours on the handoff you know how do you receive you know someone's coming you start running you receive the baton or you hand the baton off a certain way you take off because if you drop the baton you're out of the race your team's out of the race your school's out of the race your country's out of the race it's a big deal we have the same scenario in biomedical sciences we need to go from discovery to validation integration to practice integration to policy we can't one person can't do all this these to be people who are really good at this and really good at this really good at that but our handoff is lousy often my dissemination strategy is osmosis you know I publish something and hope that someone accidentally reads it as opposed to saying alright who's going to use this data how do I interact with them so I make sure that the data we're putting out can be most useful to them as they run forward you know walk through any foyer or any atrium in any research biomedical research institution in the United States and there are batons all over the floor you're going to trip on a baton because we've dropped so many of them over time but we have to get better at that and I think that's the big challenge now if we want to make progress yeah we need to be smart yeah we need to have the latest technology but we need to be thinking about how do we do this relay race and how do we do it better and it means we have to talk to people we may not even like and make sure that they're on board and that they're ready to receive what we're doing because otherwise we might get promoted we might get a free trip to Bethesda but are we really going to help a single person and so that's the challenge not just for pharmacogenomics but especially in this area you know how do we do it better and I'll stop at that point thank you very much any general questions and then and then anybody who once can join the conversation hi uh may I ask a question hi excellent talk fascinating subject I had a question about the complexity of the challenge of pharmacogenomics I had a the pleasure of spending a little time with the PGRN about a year and a half ago this NIH supported pharmacogenomics research network and the the issues that seem to confront it now particularly in the area of cancer therapeutics have gone up almost exponentially in complexity with the recognition that cancer tumors particularly the solid tumors are not homogeneous genetically and so I wanted to know what your thoughts are about how this is going to impact pharmacogenomics well very much so in terms of the cancer side I mean now what we're doing so before with next gen sequencing of tumors clinically you thought oh maybe we need to do 30x coverage 30 times coverage for that we're now doing routinely about a thousand x coverage and the reason why is that we know there are subpopulations that are there and you can't see them if you just do 100x or you need to do very deep resequencing clinically in a clear environment in order to find those and act on them accordingly we're also using serum more often now or plasma to look at our mutation clones mutants that are rising over time kind of in a minimal residual disease type model that you'd use in leukemia because we can with the technologies out there one can now find things way before there's an imaging change or there's a symptomatic change and decide whether to act on it and so it's it has changed our complexity quite dramatically but there are some institutions like our own that have really embraced this and are trying to use the heterogeneity that is definitely present to as an advantage in terms of choosing therapy it is there's no doubt that life so life never was simple it's just that we like to think it was and in some ways we're just kind of hitting reality a little bit harder than we want but you are exactly right that whether it's an autoimmune disease or a complex cancer life is not as simple as it used to be another quick question sure if I may do you see the benefits now of stem cell technology in terms of setting up high throughput screenings for both cellular and neuropathic toxicities as a pre-event before going to say humanized animal models and human yeah so I'm still a little bit skeptical about that approach for the treatment of cancer in cases where there's so much heterogeneity because you'd have to do so many things you'd have to introduce so many I was referring to Marker and oh Marker oh yeah okay because we are looking at stem cell therapies for toxicity you know if you can introduce stem cell therapies to someone who has a genetic predisposition and skipped some of those slides for neuropathy one can try to bypass that event that's still very much a research tool but one that is evolving in terms of markers of stemness or whatever you want to call it at the moment they are definitely measurable or at least there are measurements that are called stemness but we haven't figured out we haven't figured out yet how to apply them in a regular basis clinically so they're very much a research tool there are people now that are you know flow sorting out at many institutions including around you know flow sorting out these different sub types of cells reintroducing them into mice and other systems trying to understand does one is that really a subpopulation needs differential treatment how does one treat it it could be that once we're smarter we will treat that small subpopulation and ignore the rest but at least that's the way I'm not sure I answered the question the way you meant it but you can kind of see the direction things have been heading