[MUSIC] For the last module, I reviewed where we have been and the kinds of concepts that are popular now and that are being worked on in personalized medicine. I want to continue that theme and look a little bit toward the future of where this is going in the next year or two or three. I think it's probably arrogant to try to think of where we're going to be in ten or twenty years, but here are some areas where people are working on them right now. I don't think they're ready to be implemented in any kind of clinical personalized medicine space, but they're really interesting and they may be ready soon. So one of the areas that we hear about a lot is the so-called microbiome. When I was in medical school, we always learned about the bacterial flora that populate our bodies, usually in the gut, but it turns out that they're in many other places. And this is a cartoon from a recent review article showing that there are normal flora that inhabit almost every part of our bodies and they're made up of different types of bacteria. And the recognition that there are probably for every DNA molecule that comes from our germ line, there are probably 100 or 1,000 or 100,000 DNA molecules that come from the bacteria that live with us. Had led people to think about what do those bacterial do and how are they influencing the diseases that we get? So one very obvious place to think about that is in gastrointestinal disease, because we all have the GI tract is full of bacteria and people have looked very, very hard at the relationship between the microbiome and inflammatory bowel disease on the left or gastroesophageal reflux disease on the right and there are associations. And the hope, of course is that by adjusting the microflora, you can prevent the microbiome, you can prevent these diseases or at least have an impact on the rate at which they occur. Now those are obvious examples, but there are other examples now where there's increasing evidence where the microbiome may play a role in In disease susceptibility. Asthma is one or two on the upper left. Psoriasis and then obesity. And most generally, the questions around the microbiome are how do you get the microbiome? How is it populated at birth and does it change after birth? Do early antibiotics predispose to obesity? There's a nice line of investigation that suggests that. And is there a connection between the microbiome and changes in the microbiome and other diseases who's incidence appears to be changing quite rapidly? Food allergies, peanut allergies, asthma, diabetes, obesity, inflammatory bowel disease, even autism. Is there a relationship to cancers in the GI tract and elsewhere? What happens when you feed people antibiotics, when you feed animals antibiotics? How does that affect us? Does the microbiome change the way in which drugs that are taken orally are absorbed? And can we use the microbiome to look at health? Can we use the microbiome to develop better antibiotics? So those are very, very exciting areas of research that will probably have an impact on the way in which we personalize medicine over the next half decade or decade. Another area that's really exciting is illustrated by these two examples here. This is an example of genetic changes in schizophrenia and the problem of copy number variation and then the problem of the extreme QT prolongation in autosomal recessive case of the long QT syndrome. The problem with both of these diseases is that investigators would be very, very interesting in studying, for example, brain cells from a patient with schizophrenia and these variants or heart cells from a patient with the long QT syndrome due to these variants. Studying the individual cells from the affected patient. Now, it's not possible in general to accumulate brain tissue or heart tissue. You can do it by biopsies, but those are very, very specialized circumstances. So within the last half decade has come along an ability to take a skin biopsy in a patient with a known phenotype. That phenotype can be schizophrenia, it could be a cardiomyopathy, it could be cancer susceptibility, it can be a heart arrhythmia disease. It create fiber blasts in a dish, coax the fiber blasts to become pluripotent stem cells. Coax those pluripotent stem cells to become the tissue of choice. Neurons, cardiomyocites, intestinal cells and then study the effect of disease in that particular patient in the test tube. So that's a very exciting area, stay tuned. Now I've shown you this slide before, the idea that we're all descended from out of Africa populations and the idea again that we are different based on continental ancestry, whether you're European, African or Asian. And also within those continental divides within those continental divisions, you can actually subdivide populations across geography. So this is a very nice illustration I've shown before of how the purple and blue parts are Iberian peninsula genes and the pink parts are Central European genes and the yellow parts are Greek genes and they're different enough that you can actually identify some populations within a European ancestry population. At that comes along with increasing genetic mapping of common and rare variants. And ultimately, from sequencing. So that's an interesting area, because if there are diseases that are unique to these populations, we can study those diseases enriched in those populations. Understand the genetic mechanisms of disease and then generalize those genetic mechanisms across the world. Here is one example of how important ancestry can be in terms of traits that we don't ordinarily think of as genetic. This is the distribution of reference or wild type alleles and varied alleles color coded as you see in the lower left-hand corner in a gene called APOL1. APOL1 is thought to confer the APOL1 variance are particularly prevalent in West Africa and our thought to confer protection against sleeping sickness and that's why they persist in the population. Now many of the African Americans, most of the African Americans who live in the United States have their ancestry in Western Africa. So this allele happens to be quite common in African American populations with minor allele frequencies of up to 25%. There are actually two variance called the G1 and G2 variant but they come from the same kind of phenol types. And of course, sleeping sickness is not a big deal in the United States, but what is a big population in the African population in United States is hypertension and diabetes. And the problem with hypertension and diabetes is that many of those patients go on to get renal failure and what's interesting is that these alleles appear to predispose patients to getting more rapid development of renal failure. More rapid development for the need of dialysis, more rapid development of the need for kidney transplant, if they have hypertension. And this is one of many articles that discusses that. These alleles are not present in the European ancestry populations or the Asian ancestry populations. So this is an example of how it becomes very, very important to study many, many ancestries in order to get a picture of what the physiology of disease. And the pathophysiology of disease progression, disease development, disease drug responses might be. Another example is an interesting issues that revolves around the Havasupai Indians and their foray into genetics. The details of the story are controversial, but what is clear is that investigators went to the Havasupai because they have a high incidence of type two diabetes. And with the engagement of the tribe, started a project to look at whether they could identify predictors of type two diabetes among the Havasupai Indians. They collected phenotypic information around diabetes and they collected genetic samples. Now the problem became that it looked like the genetic samples might have been used for other uses. And there was a big controversy around that and it included a monetary settlement to the Indian tribe and ultimately destruction of the genetic samples. And the story is important to tell not because it's somebody's fault or not because it should have been done this way or that way. But it raises very, very interesting issues around the use of DNA samples in populations that have confined or restricted genetics. So a population like the Havasupai are an isolated population and they may have genetic variances that don't exist or don't exist at high allele frequencies elsewhere. So one issue that was raised is the issue of informed consent. If you consent to a diabetes study using GWAS, should your samples be used for a diabetes study for sequencing. Could your samples be used to study a different disease? Could your samples be used to gauge the frequency of alleles without regard to disease. And each of those raises interesting questions about what exactly people consent to and what exactly people will consent to in the future. So samples that were collected ten years ago for genetic studies might have included the possibility of studying other diseases but probably didn't include the possibility of whole genome sequencing. Does that mean you can use those samples? Probably not unless you go back to those subjects and ask them again. What do you do with the results? We'll talk about that in a moment. The Havasupai data were also used, or could have been used to study patterns of human migration of the type that I showed on a slide earlier. But the problem there is that the lore of the tribe may have particular value to members of the tribe. And they may not be interested in understanding the genetic base, using genetics to study human migration. They may have their own lore and they don't want to have their genes used to study that particular problem. They might want to have their genes used to study diabetes. If you find genes that predispose to obesity, or predispose to autism, or traits like that, the tribe, or individuals within the tribe could be stigmatized. The question of how do you go about studying diseases requires that you have the genetic information, plus access to electronic or other kinds of medical records. And there's always a question of how that's going to work. And if you have a large data set of patients, there are ways in which you can strip those records of identifiers but they are never perfect. And so there's always this potential that if you're part of a genetic study, your information could become available publicly through malicious attack on the records. So the Havasupai issue raises all kinds of things that we have to think about as we go forward in genetic testing at a clinical level and genetic research using large scale sequencing. So I've shown this slide before showing that sequencing costs are dropping dramatically. They seem to have plateaued in the last year or two, interestingly, but there's ever hope that they will drop again with the advent of new technologies. And within 5 or 10 years, we might even have the $100 genome. And of course you don't know what to do with the $100 genome, it's the million Dollar informatics problem attached to the $100 genome, that's the real problem. So, what are we going to do is not very clear, but here is a really interesting paper. And it's actually the first example of a full sequence human genome that was annotated for all the variants that might be of interest. The story is that the author whose name is highlighted, Stephen Quake, was the developer of one of the newer technologies to sequence human genomes. He sequenced, of course, his genome, and then he went to the informatics group at Stanford, and asked, what do I do with this information. And they spent really a couple of years annotating his human genome sequence, in three teams. In a team that looked at common variations, GWAS kind of variation, and so often they were increased or decreased risk. If he was at increased or decreased risk of common diseases like diabetes or coronary disease. They looked at a set of rare variants, we'll come back to that one in a second, and they looked at pharmacogenetic variants. And at the end of the day, the pharmacogenetic variants may be the only one of interest right now, but going forward, that may change a lot. So at the time this paper was published, he had 2.6 million SNPs, now he probably has more than that in fact. But the human genome annotations weren't sufficiently dense to actually understand what's a variant or not. Lots of copy number variations probably has more than that and when this paper came out in the late press, it was really very, very interesting. There was a focus on pharmacogenetic variance, because those seem to be actionable right now, at least some of them. And then there was a big focus on whether he was an increased risk for coronary diseases, and he was 40 at the time, should he take a statin? And it had to do with the fact that he has a couple of GWAS related genetic variance that increases risk for coronary disease quite modestly. That is not a question where genetics are going to probably play a major role in clinical decision making, there are other ways to do that. The other problem is looking at rare variants. Now I've shown you this slide before. And I'll show you some of Steve Quake's data in a moment, but in this slide which I've shown you before, highlighted the fact that this was a study looking at Framingham and the Jackson Heart study. Framingham largely European American, and Jackson Heart study used mostly African American. 11% of 3,600 subjects had rare non-synonymous variants in eight hypertrophic cardiomyopathy or dilated cardiomyopathy genes. That's a huge number, though does that mean 11% of the population has those diseases? Probably not, because those diseases are pretty rare. Now you can go through those 11% And annotate them using informatics tools and try to figure out whether they're likely pathogenic. So you look at things like are they conserved across species? If that particular residue is conserved across species and is now changed in an individual it's more likely that it's pathogenic. Those are the kinds of things that people can do. And, obviously, if its been reported before in a family with hypertrophic cardiomyopathy, that makes it more likely to be pathogenic. So only 22 of those variants were judged to be likely pathogenic. But out of those 22 patients, only 4 had cardiomyopathy. So we're going to have the problem of finding variance through sequencing in genes that we know can have variance that cause serious disease, and we won't know what to do with them. The problem is called the variant of unknown significance problem. I alluded to that as well when we talked about genetic testing for cancer susceptibility. So if we do BRCA1 testing, you can find a mutation that's previously been associated with cancer. You can find a mutation in a family member that has previously been associated with cancer, but you can also find a variant in that gene that's never been reported before. And you can't, with a straight face, look at the patient and say, you are at risk for cancer or not. So we have this variant of uncertain significance problem, which is going to be an enormous problem as people start to do sequencing. So this is Steve Quake's family tree. You can see he's indicated by the arrow in the middle and he has a father and an uncle who have coronary disease and aortic aneurysms, interestingly both of them, so that probably has a genetic basis. But what's interesting is his first cousin, cousin number 2, who has this disease called arrhythmogenic right ventricular dysplasia, ARVD. And what’s much more disturbing is that one of that cousin's sons, so Steve Quake's nephew, died suddenly, presumably, of that disease, arrhythmogenic right ventricular dysplasia. That's a disease where the disease genes are known. And when you look down the list of rare variants that Steve Quake has, one of the rare variants is in a gene called Desmoplakin,and he has a Desmoplakin variant that's never been seen before. So does that mean he has the disease, or does that mean he carries the same gene as his nephew, and his nephew had something else that made him have the disease, and the patient will not get the disease? Or is this a completely irrelevant finding? So even in the very first genome, whose annotation was reported, we have all the problems around this. Do you tell him this? Of course, he wanted to know everything. What do you do with this? How much do you investigate whether he has this disease? If he has the disease, do you treat him? There's not much treatment for this particular form of cardiomyopathy, but they are susceptible to serious arrhythmias and there are ways of predicting that. So the question then becomes what do you do with these variants of uncertain significance? And it's going to be an enormous problem. Here's another problem, and that is the problem of variants that predict really severe disease that is untreatable. The picture on the right is the famous folk singer, Woody Guthrie, who died of Huntington's Disease, a disease of progressive dementia that develops in people in their 20s, 30s, and 40s, and is currently untreatable. His son, Arlo Guthrie, probably does not have the disease because he has not been found to have the disease and he's old enough now. So [COUGH] sequencing raises all kinds of interesting questions. If you go out and have your entire genome sequenced, one of the questions you should ask is, well, how many variants do I have, that's an interesting question. But what about variants that actually mean something to me? The term we use, actionable. So are there actionable variants that suggest I shouldn't get certain drugs, that suggest I should get higher or lower doses of certain drugs, that suggest if I have a disease I should get drug A instead of drug B. Do I have variants that change my risk for certain diseases a lot? Do I have variants that change my risk for disease a little bit? Do I really need to know about that? Do I have anything that might affect my health? Do I have anything that might affect the way my body responds to the environment? What kind of evidence do I need to have in order to change the way I behave? Do I need a randomized clinical trial for every single genetic variant? That's probably not going to happen. Is it good enough to have an anecdote? That's probably not good enough either. Do I need to understand the underlying biology, and then have a genetic association, and together that somehow drives me to change the way I behave? That's sort of the way we do things a lot with pharmacogenetics right now. We understand the biology and we have some outcome data. So we say that certain patients should get adjusted doses of warfarin, or certain patients should have higher or lower doses of clopidogrel, or some other platelet inhibiting drug. And, then, what do you do if you find that you have a genetic variant that predisposes to a disease? Do you take protective measure against that disease? What happens if that disease is fatal and untreatable, in the case of Huntington's? Do you really want to know? Do you not want to know? And if you find out and you say, well, I don't want to know, what about other members of your family? Maybe they do want to know. Or if you say I want to know, but your sister doesn't want to know, what do you do about that? And when you think about the role that genetics plays in disease susceptibility across life, there is a group of people who think that the right time to do sequencing at the whole genome level is at birth. And that you archive the information and they use it as time goes on. So when somebody is exposed to clopidogrel, you can look up their CYP2C19 genes. But the problem there is going to be the problem of the variant of uncertain significance. We all have many of those variants, and we're just now beginning to understand the fact that they're much more prevalent, and those diseases are probably much less penetrant than we gave them credit for. And how to act on those in a sequencing environment is going to be an interesting question that the field is going to have to work on. So, I show this slide at the end of this to bring us full circle from where I started the course. And, that is, part of the personalized medicine attitude has to be, first of all, it's not only about genomics, but about many, many other things that affect what makes you a human being, and makes you respond to your healthcare needs. An ability to read. An ability to be able to take 20 different medicines, some of them once a day, some of them three times a day, some of them as needed. You may have particular cultural attitudes towards disease susceptibility or disease treatment. You may have variable access to the electronic tools that many of us are going to need to keep track of this information. And you may have personal attitudes that change the way in which you want to respond to information like whole genome sequences. So, at the end of the day, personalized medicine has to involve engagement of the individual whose medicine is being personalized, the patient, as well as a healthcare system that helps understand what those variations mean. [NOISE] [APPLAUSE]
