Scientist Stories: Eric Lander, The Human Genomic Revolution Past, Present, and Future

Good evening, everybody. My name is Tom Ullrich. I'm the Associate Director for Science Communications here at the Broad Institute, and it's my pleasure to welcome all of you here in the room. And everybody watching on YouTube to our first Broad@15 talk. The Broad Institute, as you may know, came into being in 2004. And since then, we've been witness to remarkable advances in many fields of biology and medicine. This series of talks is going to celebrate that amazing pace of scientific progress over the last 15 years and the parts that we here at the Broad Institute have played in it. It's my honor tonight to introduce our first Broad@15 speaker, Dr. Eric Lander. Eric is the president and founding director of the Broad Institute and is also— was also a principal leader of the Human Genome Project. He's going to set the stage for the series tonight by giving us a survey of the impact of sequencing the human genome, what we've learned over the last 15 years, and what lies ahead for us. Eric is delivering this lecture tonight with the support— or in memory of Eliana Hector and with the support of the Eliana Hector Memorial Fund. And it's my— I would love to make a special introduction or a special welcome to members of the Hechter family and their friends who are here with us this evening. We invite you to ask questions during the Q&A portion of the evening. We'll have microphones available. We'll also be taking questions by text message at the phone number that's been coming up on the screen behind me. And if you're watching on YouTube, we'll also be taking questions by YouTube chat. If you happen to be tweeting tonight, we do ask you to use the hashtag #Broodat15 as well. Without further ado, Eric Lander.

Well, actually, my backup one. What happened to my first slide? I don't know. Did it move? Maybe. This Broad@15 lecture series is really a way of celebrating what has been an incredible period of science. The last 15 years— maybe not exactly, we could go back a little before that, 25 years— this period of science has been extraordinary. And we thought that to celebrate the 15th anniversary of the Broad, we would ask a bunch of people people in the Broad community to give talks about the past, the present, and looking ahead to the future. It's a lot of fun to do that, to kind of step back. Day to day at the bench, one's always mindful of the fact that progress doesn't always go so fast. But if you integrate over a year or several years or a decade,— it's just stunning what has been going on. And so this is a chance for us to look. It's not only about things that have gone on at the Broad, but topics that the Broad cares a lot about and works in a lot to kind of review this whole sweep of history. One of the problems of agreeing that we should have a Broad@15 lecture series was that when the question came, would I give the first talk in the Broad@15 lecture series, I couldn't really say no. And so here I am to kick off this series. But it is a special lecture in this series because it is also, as was mentioned, the Eliana Hechter Memorial Lecture. Eliana Hechter was a remarkable person. She worked for a significant time in my lab. I collaborated with her scientifically, so this is particularly meaningful for me. And before her untimely passing, she was a magna cum laude graduate of the University of Washington at the age of 18 with a degree in mathematics and went on to become the second youngest Rhodes Scholar in history. There was a long history, by the way, of Rhodes Scholars. She then continued her studies in medical school at Harvard and at MIT. But beyond those things, she was also a remarkably gifted writer. She wrote short stories. She wrote science. And in everything she wrote, she did it with tremendous care and tremendous thought. She's the only person with whom I have ever engaged in a 20-minute discussion on the difference between in contrast and by contrast. And I learned a lot from that conversation. And so it's somewhat fitting that we, for someone who cared so much about communication, we have a lecture in her honor. In my own lab, Eliane and I had worked together with another person, a postdoctoral fellow, Orit Zook, on a really interesting exploratory paper about missing heritability of human traits. And I think very much of her when I think back to all of that work. So I wanted to particularly call attention to the fact that this lecture was special and Welcome Michael Hector, Eliana's father, and Jacob Lurie and others here today. So past, present, and future. I get to do the human genomic revolution, which has really been remarkable. I'm going to tell you a lot of science, but not require that you know lots of science start, but I'm not going to pull the punches because I'm going to really try to tell you this amazing arc that has happened. And there are going to be times I'm going to tell you really interesting, complicated things because I think that's the way to best appreciate the journey. But don't worry. It's not like people didn't wonder about the causes of human diseases. People had been wondering about the causes of diseases that ran in families for centuries. People knew that certain diseases, certain traits ran in families. But actually pinning down what was the basis of those diseases was really hard. It could be done sometimes. For example, sickle cell anemia was worked out because it was obviously a disease of red blood cells. And red blood cells basically are bags of hemoglobin. They didn't have a nucleus anymore. There's the bag of hemoglobin. If there's a problem with the red blood cell, it's probably the hemoglobin. It's a pretty good guess. And it was worked out a long time ago that it was. It was a single change in a building block, an amino acid of the hemoglobin protein. And that was worked out, its molecular mechanism. Then there's phenylketonuria. Used to be a leading cause of mental retardation. But today it's not anymore, because people realized that phenylketonuria was caused by the inability to break down a nutrient, the amino acid phenylalanine. And now every baby in the United States gets a heel stick at about 8 days of age to test and find the 1 in a million kids who has phenylketonuria so that they go on a low phenylalanine diet and don't develop that mental retardation. It's a brilliant thing. In fact, if you look on a Diet Coke can, next time you get a Diet Coke can, there's a warning to phenylketonurics. That it contains significant amounts of phenylalanine, because aspartame has phenylalanine in it. So there's even a story about genetic disease on every Diet Coke can. Well, but these are great. But the thing about both of these is that they were ad hoc. They were not general strategies for finding disease genes. They required inspired guessing. And there's nothing wrong with inspired guessing. It's just I don't know how to do it most of the time. Of the time, and one would rather have a systematic approach. So the story I'm going to tell you is how science moved on to take on diseases where the genetic cause, the protein, the mechanism in the cells were not obvious, where there were not candidate genes. When I first came into the story in the 1980s, everything was about you had to have a candidate gene. But how would you get it? And there was a big fat book, Victor McKusick's Catalog of Mendelian Diseases, single gene diseases that you could tell by the way they were inherited in families. And there were 4,000 of these things. And there weren't good candidates for more than a handful. And then what about things that weren't single gene diseases that transmitted in families like Mendel's rules? Height, weight, schizophrenia, heart disease. How in the world are you going to find out, well, it's not a single gene, multiple genes that might be contributing. We didn't even know how many genes might be contributing to that. I'm just going to tell you the story of how that happened, where we are today, and where we're going. Now, to do that, I'm going to briefly take you back not 15 years, but a lot earlier than that to 1865, because if we're going to talk genetics, I have to at least mention the founding of genetics in 1865 by Gregor Mendel. Gregor Mendel was a remarkable scientist. And they always tell the story in the high schools about Gregor Mendel was this monk working in a monastery who did these experiments with they don't tell you he actually studied math and science, math and physics, and was hired by the abbot of the monastery who was basically running a small biotech incubator in the city of Brno. And they deliberately were working on this stuff in order to improve agriculture to make the world better. So this was not some isolated scientist. He was in kind of the Cambridge, Massachusetts of his time. Time in some ways. Now, he worked out this beautiful mathematical model from which you could infer that there must be particles of inheritance, what would later come to be called genes. And it was so beautiful and so mathematical and explained so much that it was completely ignored for the next 35 years because because biologists don't like abstract mathematical things. If you're inferring the existence of genes, they say, I want to see the gene. Where's the gene? What is it? And an abstraction doesn't really help anybody. So it largely went dormant for about 35 years until the year 1900, when genetics was rediscovered by 3 different groups simultaneously in the year 1900. And this field got underway in a serious fashion. Studying genetics, studying mutants, studying organisms that lack a particular gene and that causes a particular trait, what we'll call a phenotype. Now almost exactly at the same moment, really just a couple of years earlier, a parallel field got started— biochemistry. Grinding up organisms and purifying a particular component that could carry out a function, like a chemical reaction, digesting sugars, grinding up yeast to figure out how can they digest sugars, and purifying components. And these two complementary methods, studying an organism minus a component, studying a component taken away from the organism, were almost perfectly dual to each other. They were both powerful. And for almost 50 years, they had almost nothing to say to each other. Because nobody understood what those components were. And then the really interesting intellectual advance in the 20th century was the recognition that they were connected. The components the geneticists studied, genes, actually encoded the instructions for proteins. And that recognition that these were two sides of the same coin really gave rise to this field of molecular biology. The recognition that DNA, its double helical structure, let it replicate, that it encoded the instructions for proteins. And we now had this amazing triangle, this intellectual unification. And by the 1960s, people had even worked out the genetic code. They knew which 3 letters of DNA specified which building blocks of proteins, a lookup table of 3-letter codons to amino acids. They were so pleased with themselves that they declared victory, said that molecular biology was now largely solved, and let's go on and study the brain. And people did. Some people left the field because they figured it was done. And then, as so often happens, a younger generation arose. And the younger generation said, whoa, not so fast. We actually can't even read a single we can't read a single gene yet. We know in abstract what the genetic code is, but we can't read one gene. Maybe we're not finished yet. And they gave rise to amazing technologies in the 1970s of recombinant DNA. The recombinant DNA revolution gave us the ability to purify genes away from each other by putting them into little pieces of DNA that could replicate in bacteria. And each bacteria would pick up a different piece of human DNA and copy it. And then you could get purified forms of individual genes. And then they worked out how to study those genes, and even to the level of reading out their DNA letters so that you could start sequencing genes. And you could use that to find the gene where you knew the protein— insulin. If you knew insulin, you could find the gene for insulin. You could clone the gene for insulin, make tons of insulin, and give it to diabetics. If you wanted to find the gene for growth hormone, something that had been discovered, you could look for the gene for growth hormone that encoded that particular protein and produce recombinant growth hormone, and hemoglobin, and onward, as long as you knew what you were looking for. And yet, to come back to what I said at the beginning, there was a problem. For most diseases, we didn't know what we were looking for. So while this technology was super cool and led to the entire biotech industry, it could produce recombinant proteins, but it couldn't find the gene for cystic fibrosis because you didn't know what you're looking for. And this is the story I want to tell. The principles first for how you could map genes when you didn't actually know what you were looking for? It's not so— I mean, sometimes it's hard to find something when you know what you're looking for. But it's much harder to find something when you don't know what you're looking for because how would you find it? Well, the answer was that a brilliant MIT professor and colleague, David Botstein, in 1980 proposed an extremely simple, elegant idea. That had actually been developed in fruit flies back in 1911 and had been used for fruit fly genetics and corn genetics and other things, which is that when genomes, when chromosomes are transmitted in families and across fruit flies or across maize or other organisms, things that are nearby on the chromosome tend to travel together. Sometimes they genetically recombine, but the closer they are, the more often this form of the gene and that form of the gene that are bound on the same chromosome would continue to travel together. And so in fruit flies, what they did was if you had a trait that you didn't know what the gene was, you could look at how it was inherited in a family of fruit flies relative to lots of other genetic markers. Genetic markers that controlled white eyes, curly wings, funny hairs, other things. And they built genetic maps of all these funny traits in the funny flies. And then they mapped the new thing relative to that. The problem with doing this in humans was we haven't got wings. We haven't got these funny white eyes and other— and also you're not allowed to set up crosses. I mean, you're allowed to set up crosses with somebody you're— you have to talk about it and all that. And then it turns out you never actually produce a statistically significant number of children, because those of you who have children know that reaching statistical significance is not compatible with raising the kids. So the question is, how are you going to do this for the human population? That's what David Botstein worked out. We don't need to worry about it. Just take natural families. There's inherent genetic variation in all of us. And let's just pretend that all that genetic variation we put there deliberately and just use it to trace inheritance in families. And so he said, find lots of genetic variation and trace inheritance with it. It was a brilliant idea. And it actually worked. I met David in the early 1980s. And we began talking about, well, what would you do if you didn't have a single gene trait and you didn't have families? Could you do this even if you didn't have families and the genetics was more complicated? We actually worked out ways to do this for common diseases that might be polygenic, might have multiple genes, by instead of tracing things slavishly in a family, treating an entire population as if it was one big family. And using lots of genetic markers to recognize segments of the genome that had come from a common ancestor, common ancestral segments. So you could pretend like it was a big family. I just happened to have lacked the previous 100 generations. But if I had a lot of genetic markers, I could still see that. Now, this is all great. It was used in the 1980s to clone a couple of genes, the cystic fibrosis— I remember the day David Botstein crossed the street, called me into his office, closed the door, locks it, and says, the gene for cystic fibrosis is on chromosome 7. He was really excited about this. This was pretty cool. Now, at the time, the genetic marker that was linked to the cystic fibrosis gene, it traveled together 90% of the time, which sounds good, but it's actually terrible. Because it meant it was 10 million base pairs of DNA away.. And in those ancient days, traveling a distance of 10 million bases to find the gene was really hard. Anyway, it took 5 years of work. It took tens of millions of dollars, probably 100 people working on this. And by 1989, they found the gene for cystic fibrosis. And when they found it, it looked like this. It had a lot of As, Ts, Cs, and Gs. But I call your attention to the little red box there, because those 3 letters, TTT, were missing in most of the people who had cystic fibrosis. A deletion of 3 letters deleted 1 building block, 1 amino acid, phenylalanine at position 508. And that was a major cause of cystic fibrosis. Not only that— so that means you could do genetic testing and see who's got that. It'd be very easy. Not only that, you could toss the sequence of the protein in the computer and ask the computer to compare it to every other protein anybody else had sequenced to that date. And ask, was it similar to anything? And they did that. And they found, oh my goodness, it's similar to a whole bunch of other proteins that sit in membranes and transport things. You must— the cystic fibrosis gene must be a transporter of ions, in fact. That's amazing. You didn't have to do any experiments. You sequenced the stuff. You could do genetic diagnostics. And you could find out— that the function was probably this. You have to do an experiment to confirm it. But being handed a hypothesis on a silver platter was pretty amazing. So this was great. Everybody was very excited that we could find genes without knowing what they were. There was only one problem. I mentioned lots of years, lots of money. Lots of people. And that was just one of thousands and thousands of problems you'd want to apply it to. The problem is the human genome was a very big place. It was 3 billion bases long. And a typical gene might only have 3,000 letters of coding information. And the mutation might be 1 or 2 or 3 letters. And how would you find this needle in a haystack? Actually, if you do the calculation correctly, it's somewhat worse than a needle in a haystack. Needles make up more than 1/3 billionth of of a haystack if you calculate correctly. So somewhat worse than needles in haystacks. Well, that is what led us as a field to the idea that we needed to actually map the human genome. We were going to have to make this much easier to do. So a plan was laid out. If we were going to have these genetic markers to trace inheritance, we'd build a genetic map of lots of genetic landmarks. We then needed to fill in all the DNA in between so we could cover that distance. So instead of having to tediously cover the distance, We could just look it up. We would have the whole sequence. For cystic fibrosis, they spent a long time finding the region, sequencing it. We want to just be able to double-click on it, and up pops the sequence. We want to have a list of genes. And importantly, we want that to be completely freely available to everybody around the world so anybody can do stuff like this. The Human Genome Project got going. It got finished. I'm not going to say that much about it. I lived through all of this. In fact, Here, the Broad grew out of the Human Genome Center that existed here in Kendall Square in a beer and popcorn warehouse not far from here, just a few blocks away, where the human genome got sequenced— where about a third of the human genome sequence was done. But it was an international collaboration involving groups in 6 different countries, 16 different centers, everybody putting their data freely available on the web every day. Yeah, it cost— $3 billion. But thanks to the American Congress and the foresightedness of the government saying we should build these things as public infrastructure, it got done and made freely available. And $3 billion turned out to be only 1% of the National Institutes of Health budget. So it sounds like a lot, but when you divide it over the 15 or so years and you compare it to the NIH budget, it was a mighty good investment. So it got finished. There was a draft by the year 2000, a finished sequence by April 25, 2003. That was not an accident. That was 50 years to the day after the Watson-Crick paper. And everybody pulled all-nighters to make sure we could get done by the 50th anniversary of the Watson-Crick paper. So we had a sequence of the human genome. So now the question was, could we map disease genes in a simple way, like falling off a log? And for these, remember we had this rare Mendelian, trace them in families, common disease, look across populations. Well, for the rare Mendelian diseases, the single gene diseases that transmit cleanly in families, it worked like a charm. In the 1980s, I said cystic fibrosis took 5 years, lots of people. Huntington's disease was worse. It took 10 years.. But by 2003, when the Human Genome Project was finishing, all these tools being made available meant that more than 1,700 Mendelian diseases had been solved to the extent that the gene was found. By today, 5,000 disease genes for different Mendelian disorders are known. And that's the vast majority of Mendelian diseases we know. There are very rare ones where people are still looking, and there are things still to do. But it's pretty fair to say we kind of know how to do this, and it works really well. And what used to take tens of millions of dollars, it's not a bad rotation project for a student today if you have the samples, which is what progress is in science, to see that happen. But for common diseases, it was very disappointing. People tried to do the same thing, pretending that common diseases were really a bunch of single-gene Mendelian disorders. And it just failed. Looking within families for co-inheritance— well, by 2003, there were about 8 significant discoveries across all of medicine. And even they were kind of weird, because unlike Mendelian disorders, we had single genes of huge effects. The cystic fibrosis gene, if you inherit it, well, having a single copy dramatically increases your risk of disease, because if you get another copy, you're definitely going to get it. It's like a 500-fold effect. By contrast, the handful of things that were found increase your risk by 2-fold, 3-fold, 4-fold. It just looked totally different. And it wasn't working. That was that second approach I was telling you. Look for ancestral segments that were common across a population. Collect lots of people who have a disease and ask, are there some ancestral segments that are shared amongst those people? Now, of course, to do this, you would need to have documented all the common genetic variation in the population. So let's put the refrigerator list up. Find 10 million genetic variants. Catalog the correlation structure across the entire genome of which things go with which things, be able to genotype in 10,000 people a million different genetic variants, which is like 100 billion genotypes, like 100 billion data points, and then have rigorous analytical methods. So this is kind of like insane. And so we got to work. And it got done. It got done in about 6 years. It got done because when you put your mind to it and a community comes together, we went from knowing almost none of the common genetic variations in the population to knowing tens of millions of them. We went from knowing almost nothing about the correlation structure to having an international project to define the entire correlation structure. We went from being able to genotype one genetic marker at a time to being able to do 1 million genetic markers simultaneously at a time. I leave out some details there. It was an exhilarating period, because we were building the infrastructure. We as a community here— the Broad and the Broadies played a huge role in this, but a whole community came together to do this. Also, I should say that with regard to sequencing, the cost of sequencing plummeted. That sequence I told you about, the first human sequence, was $3 billion. That is not, as they say, scalable. If you wanted to do like 4,000 people at $3 billion, it's roughly US GDP, which is just like very hard to write a grant for. So instead, it's plummeted. What used to be $3 billion I can get you, if you're willing to buy in bulk, for $600. I think it will drop to $100. That is like mind-boggling. It's 3 times faster than Moore's Law, the cost of sequencing. You want your genome sequenced? $600 today in bulk. So we now come to the situation of, can we find those genes for common genetic diseases? I can tell you, they were very slow pace of discovery. Each year, we got just a handful of discoveries that were coming along. And 1, 2, 1, 1, 1. And then the tools began to fall into place. And in 2005, 4 discoveries. 2006, 8 discoveries worldwide. 2007, 11. A lot of discoveries as all these tools kicked in. Now, you might think everybody would be really excited that we got all of these genetic discoveries that were coming along. However, scientists have an amazing ability to be grumpy. And so people said, yeah, OK, it's nice, but you know— most of the genetic effects you found have only small magnitude. They might increase your risk by twofold, or even just 50% or 20%. I'm not impressed. And they didn't land in the coding instructions for genes. They landed in the non-coding, dark matter, we don't know what it did kind of regions of the human genome. And it didn't explain an awful lot of the total heritability. And how could you interpret it anyway biologically? And so people immediately said, maybe these are just statistical artifacts, or they're something else, or maybe it's caused by rare genetic— rare variants like Mendelian diseases, even though we had, of course, found that it wasn't caused by that. But nonetheless, the pendulum was swinging back and forth. But a perfectly plausible answer was that the sample sizes were actually rather small. And maybe if we just looked at more people, we'd see more stuff. And that turned out to be the answer. Turned out a study that was done here at the Broad on schizophrenia looked at 3,000 cases with schizophrenia, 3,000 controls, and it found absolutely nothing. Not a single thing crossed the threshold of statistical significance, which if you're looking at 6,000 people and looking at a million genetic variants in them and spent a lot of money on it, you are somewhat disappointed. However, if you're mathematical, you look at the data and you say, yeah, but it's not like randomly distributed. There's still something saying— there's a signal. And it's sort of whispering, increase the sample size and we'll give you something. And sure enough, when the sample size went up to 10,000 cases and 10,000 controls, 5 significant regions of the genome were found. But it was still whispering, more samples. And at 25,000 and 25,000, it went up to 62 significant loci regions in the human genome. And then at 82,000 and 82,000, it went up to 245 regions of the human genome that were unambiguously, strongly, statistically significantly associated. And now— with larger and larger samples across more and more traits, there are more than 100,000 associations between regions of the genome and particular common diseases. 403 examples in type 2 diabetes, 273 in inflammatory bowel disease, 166 in coronary artery disease, 245 in schizophrenia. That's with common variants. And now, as people are doing more sequencing, they're finding rare variants in Suddenly the floodgates are open. It is possible to find lots of regions of the genome in the human population that have variants that are at higher frequencies in people with the disorder than not. Each of them makes a small contribution, but together they do a lot. What can we get from this? Turns out that by looking at these many, many genetic variants, It's not as easy as cystic fibrosis, where there's one gene, you toss it in the computer, and the computer will tell you, here's what this thing does. It's a transporter. You've got to look at lots of regions of the human genome. But already, discoveries are coming. Let me come back to schizophrenia. For schizophrenia, there's one region that had a screaming big signal. And I've pointed to it there with an arrow. That region lies in a critical part of the human genome that affects immunity called the major histocompatibility complex. And that led people to say, oh, maybe it's an infectious agent that's causing schizophrenia. And there's even this article in The Atlantic about how you could get schizophrenia from your cat because it would transmit this bacteria, Toxoplasma gondii. I have no evidence that any of that is correct, but it was a very amusing article in The Atlantic. But it did raise what was going on in this region of the human genome related to immunity. To make a long, long story short, investigators here at the Broad found the gene. It was really hard, and I wish I could tell you all the details because it's a beautiful story. But it turned out to be a gene with the boring name C4, complement component 4. Now, complement component 4 works in the immune system. How's it going to affect schizophrenia? Well, because another scientist working in this community happened to also be working on the complement system, she knew that it didn't just play a role in the immune system of marking bacteria to be destroyed by immune cells. It also had a second job in the body. It marked synapses in the brain to be destroyed by certain things called microglia, a certain set of cells that clean things up in the brain. And thus the idea that this was working by affecting the pruning of synapses. And sure enough, the genetic variants that were associated with schizophrenia were associated with greater pruning. Of synapses. And suddenly the pieces started clicking into place, because there had been this old observation that postmortem analysis of brains from people with schizophrenia revealed that they had fewer synapses. And the time of greatest pruning of synapses in the brain is late adolescence, early adulthood, which is exactly the time of onset of schizophrenia. And then when you start looking at those 245 genes, you say, oh, there are some other genes that are also involved in pruning, and some other things involved in pruning. And suddenly there's a hypothesis. And maybe schizophrenia may someday be treated by finding chemicals, by finding drugs that will turn down overactive pruning. Now this is just one example of that list of 245 things that are There are tons more examples of discoveries that are coming from being able to take an unbiased look to find without knowing what you're looking for. For heart disease, we all know that our lipid levels are very important. But genetics teaches us that 80% of the gene regions that have been found so far play no role in lipids. And therefore, there are a lot of other things going on than we have been thinking about in heart disease. And you all have heard that HDL is protective. Your LDL is bad, HDL is good, you might have heard. Turns out HDL is not good or bad. It turns out HDL is not the causal factor at all. It turns out something that's negatively correlated with HDL is really the causal factor—

triglycerides.

And you can figure that out. And had you known that, actually, you would have saved the pharmaceutical industry $5 billion developing drugs to increase your HDL levels.— which all failed, because it turns out it's not the causal factor. So it turns out knowing things makes a big difference in developing drugs. What? It just— you know, they figured decreasing LDL cholesterol had been so effective— everybody takes statins and things— so increasing HDL must be effective. But they skipped over the point of, was it causal or just a correlation? The genetics actually told you it was just the correlation. Beautiful work of, say, Kathreason and his colleagues here in the Broad community who demonstrated the power of this kind of thing. Alzheimer's disease. We have seen failed clinical trials for Alzheimer's disease because everybody is going after neurons. Obviously makes sense. Alzheimer's, neurons, that makes sense. But it turns out that if you ask the genetics, 80% of the genetic regions that have been found by doing these human genetic studies don't play a role in neurons. They play a role in those microglia I was telling you about, the things that clean up damage in the brain. And I think it's going to become clear that there is a neuroimmune axis where the immune cells of the brain may be the primary driving force for Alzheimer's. And you would never know except if you just ask in an utterly unbiased way and have genetics tell you, here are the things that seem to matter. Not only that, you can use these millions of genetic markers to build risk scores for people, polygenic risk scores. Who's at more risk and who's at less risk? And it turns out, beautiful work over the past couple of years has shown that we can identify 8% of the population that's at threefold higher risk for cardiovascular heart disease. Maybe they should take those lipid-lowering statins earlier. A couple percent of the population are at much higher risk, threefold higher risk, for breast cancer, and it's not the rare BRCA genes. It's a general polygenic risk for breast cancer. Maybe for such individuals, the age of mammography should be different. Maybe patients aren't all the same, and looking at their genetic profiles and correlating it for these different diseases may lead to different recommendations for different people. So all right. This is all great.

Summary—

Mendelian disease, lots of progress. Single gene diseases, we've found most of the genetic disorders that are known. The work is now on how do you cure them— gene therapies and things like that. Common disease, we know the architecture. Highly polygenic, these modest effects, 100,000 mapped. Here's the problem. I keep talking about these gene regions. And I haven't told you we know all the genes. It's because we don't know all the genes. We now have 100,000 gene regions that have been pinned down. And they're highly significant. But the work of moving from a genetic region that is linked to the disease to actually what the gene is, what cell type it works in, what process it works in, is incredibly hard. And there are heroic efforts and heroic papers of people doing this one at a time. But 100,000 is a big number, and doing this one at a time is not going to work. We are going to need to supplement those heroic efforts with systematic, comprehensive efforts, just like happened in every other stage of the story I told you. The road ahead, the road ahead is to take the same approach that has worked before. Think about the principles, build the infrastructure, make it broadly available. I will just briefly tell you in the closing minutes about these remarkable, powerful, driving scientific forces that are going to drive the next decade or 15 years. Human cohorts. Used to be you would go out and collect a family with a disease. But now, the United Kingdom goes out and gets half a million people to participate in the UK Biobank, or Estonia has 200,000 people sign up, or the United States has launched the All of Us project with a million people. Well, these very large cohorts are remarkable. And also across psychiatric diseases, in a disease-focused way, there's a worldwide psychiatric genetics consortium. I said we needed big numbers— 82,000 people with schizophrenia, 82,000 people without. Well, I believe we are moving to a world where genomics is going to get used in medicine. And the number of people relevant are everybody. That there will be a point in the not-so-distant future where it's pretty routine to simply sequence your genome. If I say it's $100, $100 is not so expensive. Amortize it over a life, around the life to 100 years, $100, like a a year. It's less than Starbucks, one Starbucks a year. We're going to get the information. The real question is, what can we learn from it, and how can we share it securely? And there are all of these different projects going on, and the world is thinking about how we're going to do this. Also, other powerful driving scientific forces. Sequencing costs, I said, would fall to $100. Then also single-cell analysis. If you keep coming to the Broad@15 talks, you will hear a talk about single-cell biology, another one of these revolutions that has gone on, which many people— but I think prominently many, many people in the Broad community have played a critical role— that are going to let us analyze millions and billions of cells and understand the genetic programs of cells at the single-cell level. And then the ability to edit DNA. If you come to the Broad@15 talks, you will also here about CRISPR genome editing and how that allows you to edit genes, activate genes, inhibit genes. And suddenly, you have all these tools. Read genomes. Study the expression patterns of single cells. Edit the genome. Tweak the genome. And then finally, data. Data science is critical. Doing this, as has been done up until very recently, where all the data that you generate is on your computer and somebody else's data is on their computer,— it's just not going to work anymore. The data sets are too big. You can't even download somebody's data. Happily, we have cloud computing. And figuring out how these data can live in clouds, be accessed by people, it's a serious high-class problem. And the Broad and scientists here are very much involved in that. They developed a framework called the Data Biosphere for making open-source software with open APIs so things could all be connected together. We helped stand up, together with others in the world, the Global Alliance for Genomics and Health to set standards. And a data platform called Terra is getting built that is, again, open source. And hopefully, the world will be able to exchange and use data, again, with real security around it. What's amazing is the surprising power of these data. Human cohorts. I could go on and on and on about how much you could learn by studying human populations. Not just who gets diabetes, but the progression rate of diabetes, the response to drugs for diabetes. All of these are traits that are affected by genetics. You can even study behaviors. UK Biobank— because it's the UK— asks all these questions like, would you call yourself miserable? And so there's a miserableness index. Are you a morning person? They have like, what do you eat for breakfast? The UK just like went to town asking questions. And it turns out you can find genetic correlates for all these interesting diet preferences and things like that. Also, the population is so wonderfully diverse. Finland is a genetic isolate where frequencies have been scattered high and low because they went through a little bottleneck. Other places, Pakistan and India. Have cousin marriages. And you see genes coming together. And you can see recessive traits appear much more easily. You can learn so much by looking at these plots I was showing you of genetic association across the genome. You can figure out, for a given gene, what different traits does it affect? You can figure out, for a given set of traits, Schizophrenia and bipolar disease— manic depression, they used to call it— do they have a similar foundation? Do they correlate with each other? And you can correlate their genetic factors and ask, are there common processes? You can figure out which cell types matter by looking at the genes that are coming up and asking, which cell types show those genes? And onward and onward. And it just turns out that when you really get at this amount of data and you correlate it with all these other tools, you can learn so much. And this business of these regions that are found in non-coding parts of the genome and not knowing what the genes are, that's just plain unacceptable. We need a total lookup table for every genetic variant in the human genome, every region of the genome. Is it a regulator, and what does it regulate, and which tissues? We just want one big lookup table. And for every genetic change, does it affect that regulator affecting this gene in what tissue? And there are amazing sets of tools that make it possible now. It's not easy, but 5 years ago you would say it was impossible. And today you can say, I see a path using sequencing, and single-cell analysis, and CRISPR, and other things to make that lookup table. And today, the things that are heroic, like with everything else we said, I I expect 10 and 15 years from now, students will take it for granted that people have always known the complete lookup table of what regulates what in the human genome, just like everybody's known the periodic table or everybody's known the map of the Earth. Students are wonderful because they so quickly adjust to what the world is and assume it's always been like that. And our job is to create that world where all of that information is available and we can start looking this up. And for that reason, the community, the genetics community is coming together around a new revolution in human genetics. The timing could not be better. Last summer, a group of folks met in New York to say it's time again to reset the vision and reset the projects. And then a meeting was followed up in Oxford. And 100 people came to that. And then a meeting was held in Helsinki. It was decided to create an International Common Disease Alliance spanning continents and countries and groups working all the way from the basic human cohorts to the cellular mechanisms to the people in pharmaceutical industries. And its first official coming-out party will be in Washington in September of this year. And we're seeing the same thing happen as has sustained each of these revolutions. It's a pretty remarkable thing. The human genome, human cells turn out to be this incredible library of information. This is the beautiful library in University College Dublin. And this is how I think about the genome and all these genomes. If you ask the right questions, you can learn so much. You can learn about disease and health. You can learn about evolution. You can learn about physiology. You can learn about migration. It's all there if we can pull out the right volumes and learn to read it in the right ways. So, that's the adventure story for today. When we said there was a sweep of history over the past century, really the last 15 years when this has happened, where we stand today and where this is going, it's really fun to share with everybody how science changes. Day to day, it is frustrating. You don't feel like this day to day. But as long as you stand back, like go home for weekends, see your family, things like that, and you stand back and you think, wow, where were we like 2 years ago? And you realize mind-boggling change happens. Nowhere did I say we're curing all disease yet. We're not. But for the first time, we really can find the causes of all these Mendelian diseases. We can find— So many of these things affecting common diseases. I have no doubt anymore that this will become the library of information for people who now go forward to create therapies. We see enough examples to know that this will become routine. It's not going to be easy. It's going to take a lot of hard work. It's going to take a generation. But I feel enormous confidence in it. So it's great fun to share it with you, and I would be delighted to take questions.

Thank you very much, Eric. The floor is open. We have folks who are going to be running around with microphones, so just raise your hand and one of us will come to you. And if you can get the text message— yep, the text message number is back up on the screen. Oh, yeah. So you can also text a question in as well.

Thank you for a wonderful, very interesting talk. I have two questions, if you have time to answer— to address both. The first is, you mentioned the effects of some occurring from the non-coding regions, and I'm wondering to what extent they're still appropriately called non-coding and what the implications of that are in terms of all that you've discussed. And the second question I believe, but I'm not positive, correct me, that the way the DNA spirals, the direction, makes a big difference, and that if it spiraled the other way, for which I believe we don't have any examples known today, that that DNA won't interact properly with any current DNA, and how does that interrelate?

Good. Two quick answers. It is non-coding DNA because coding DNA means encodes just the instruction for the building blocks of the proteins themselves. But non-coding does not mean non-important, because this non-coding DNA encodes the regulation— when that gene is turned on, in what tissues, at what times, in response to what. And it turns out, if you look at evolution of mammals, the non-coding DNA is conserved across evolution. And of all the conserved stuff over the last 100 million years, 3/4 of it is in the non-coding region. So we used to think the most important stuff, or the only stuff evolution cared about, the only functional stuff was the coding regions. No. Turns out all this non-coding stuff— not all of it, but pieces of that non-coding stuff are very important. With regard to the second question, which was on— Spiraling. There is a— we have standard B-form DNA. There is something called Z-form DNA. You're absolutely right that it's very different. Wouldn't be recognized by any of these proteins. You can make it happen in the laboratory, but it doesn't really much happen. Although I'm sure somebody will tell me that there are instances where it really does happen in biology, but it's not the norm. And so once something gets going, everything sort of locks in around it. Maybe it could have been that life could have started using that form, but it didn't, and we're stuck with it.

That's pretty good. [Speaker:AUDIENCE] So all these databases that are being created and we look to create, do they include the non-coding effects, non-coding region effects?

[Speaker:PROFESSOR] To the extent we know them, yes. And it includes epigenomics, things that sit on top of the DNA. And in a longer and more technical lecture, I would tell you about all the kinds of such information that we laid on top of it.

Yeah, absolutely. Other questions?

Let's see one down here. [Speaker:AUDIENCE] Given how much we've learned about 245 different, as you say in schizophrenia, genetic variants that you've seen, is the way we think about developing pharmaceuticals totally inappropriate in terms of one pure chemical with the mechanism of action? Is this really the right way that medicines ought to be developed? Well, I think what these 245 things are telling us is there are a bunch of pathways. Like maybe of the 245, 24 of them are working in some process. We want to understand those processes. So in another disease I know a bit about, inflammatory bowel disease, Ramnik Xavier, another core member of the Broad, has beautifully shown there are at least 10 processes. That are affected by these IBD genes. And by understanding the process, you say, can I make a medicine against that process? You don't have to make it against the gene. You make it against the process. And he's shown that you can do such things. And it may be that you want to affect multiple processes simultaneously. So really, having all these genes— people who say, oh, this is terrible. We have all these genes. Now it's great. We have all these genes. Because we can figure out pathways and processes, and that's what you really think about tweaking.

Well, I'll toss one in that's coming from the internet about sequencing— Oh, what's the internet say? The internet's curious about sequencing technology and the direction that that's going in, and is the way that we're doing it now the best way, or are things going to get better?

Oh my goodness. No, no, no, no, no. Sequencing technology undergoes revolutions very frequently. When I first got involved in biology in prehistoric times, you know, you had two pieces of glass with a thin layer of polymer, and you had radioactively labeled DNA fragments, and you pipetted them very carefully behind the plastic shield, ran them through an electric field, pried off the layer of glass, wrapped it in Saran Wrap, put a piece of X-ray film on it, put it in the freezer for a couple days, developed it, and sat there with a magic marker. It's gotten better. It's gotten better to the extent of having automated sequencing machines by a cool technology that ran DNA down a capillary, which was the coolest possible imaginable thing until it was completely displaced by another technology that involved synthesizing DNA on little strands of little pads on a glass slide. But then there's cool new technologies that involve little pores in membranes. And single DNA molecules flow through these pores. And as they flow through, the different bases slightly affect the connectivity— the connectivity— the electric currents across these little holes. And by interpreting the way that the conductivity is affected, You can infer the DNA sequences. And there are kids in labs and garages who have new and crazy ways of thinking about how they're going to sequence DNA cheaper and cheaper. I said I think we're going to get to $100. I actually know people who want to think that's too expensive and want to get to $10 for a human genome. And whether those ideas will work, I don't know. But there are a lot of things on this planet, and they all should get sequenced. And— well, because there's so much information. Tell me what you wouldn't want to sequence. Just let me know what ecosystem in ecology you wouldn't want to sequence, what evolutionary problem you wouldn't want to know the answer. If it was free, wouldn't you want all that information? And I think that's increasingly the view people are taking, that it's not like that DNA is the best way to solve a problem, but it's the cheapest way. And you start there. And it turns out if you have a lot of it, you can solve a lot of problems, surprisingly. I see another hand in the front. Question up there or there? Oh. [INAUDIBLE] What are your thoughts about regulating the misuse of this technology? What are my thoughts about regulating the misuse of this technology? It's incredibly important. I mean, there is a level of trust that must be maintained in our society about misuse, about ensuring that we do not misuse this technology. Technology, that genetic information about an individual is not used to discriminate against them, to turn them down for insurance, to stigmatize them. There are more complicated questions about groups that have their own creation stories and really aren't interested in hearing our creation stories that we would tell as scientists. I think there's a level of respect for human dignity that has to underlie all of this It has to underlie IRBs in any of these studies. It has to underlie the protection of genetic information, including the right of people not to participate in this, the right of people to share if they want to share and not share if they don't want to share, the complications of the fact that whatever I learn about you already tells me something about your first-degree relatives, because they shared half your genomes. Geneticists spend a lot of time thinking about it, and one could have a whole lecture on it. But it is impossible as a human geneticist for a field that was born in the early part of the 20th century with some of its early founders participating in anti-immigration debates to say that genetics would tell us that Jews and Italians were inferior and that immigration laws should be passed to prevent them from coming in. And then, of course, the extraordinary and horrific misuses of the Nazis of genetics. This idea that we are only this far from slipping into eugenics is something that is on the mind of any responsible geneticist. It doesn't mean we shouldn't be doing this stuff, because the potential to produce cures and the potential to produce important knowledge is tremendous. And we can't not do this. But what comes with the territory is a responsibility for the fact that There will be people who will misuse this, and we can't say we're going to prevent misuse. We are going to have to constantly police and fight misuse to make sure that we get the benefits, because they don't come without that kind of attention and policing. Yeah, question up there? Oh, over here? Oh, sorry. Thank you. Hi. You mentioned Romnick Xavier's work. Could you speak to what you think the role of the microbiome will be and how important it will be in the near future? Oh, the microbiome did not get any attention today because I, in a kind of prejudiced way, spoke about the human genome as the genome of our human cells. But of course, your body is covered with microbes. It's got them in your gut. It's got them on your skin. It's got them everywhere in different microbial communities in different parts of your body.. And they all have their own genomes. There are as many microbial cells in your body as, quote, human cells. And they come from thousands and thousands of different species. And people are reading their genomes, too, and asking, how do those communities change in response to disease? And what if the zillions of small chemical molecules they make mess with you? Well, we know they make things that mess with your immune system. In certain ways. They make things that may mess with behavior. We know they mess with obesity. Some of these may be parasites, and some of these may be commensals. Running a human body without your microbiome, it doesn't work. And so understanding which of these are actually part of the operating manual of a healthy human body, and which of these things then predispose to disease, is an entire frontier about which we should have an entire separate lecture. But I wish to acknowledge that the human genome of which I spoke was only this kind of more conventional human genome we pass on, but a richer view of the body would consider those genomes as well. We have a rare disease patient advocate who's asking this question on Twitter. I believe we are in the golden age of genetic discovery, discovery, when do you believe we will reach what historians might call the golden age of cures? So I think in the case of rare genetic diseases, we can see that we might be beginning to enter that era. To do that, we need to know how to reliably replace genes by delivering genetic cargo to the right cell types. That genetic cargo might be a virus that carries a functional copy of a gene to a patient who's missing it. And then you have to ask, how do I get that virus to the right cell types? If it's a certain set of blood cells, and all I have to do is hit 10% of the blood cells, that's one thing. Not so hard. If I have to replace a gene in many, many brain cells, how do I deliver it to those brain cells? How do I do it safely? Safely. Well, there's an explosion of methods for delivering genes or editing genes— CRISPR I referred to. There's an explosion of methods with viruses and lipid particles and other things for delivering to cells. You can begin to see the outlines of a general toolkit for being able to edit and replace genes in different places in the body of a patient or be able to deliver to different places. So you can begin to see this generality developing. I don't want to be Pollyanna about everything. All of this is hard, and some things will be much more fixable than other things will be, and some things may never be fixable, which is also why the— making available genetic screening to parents so that they are aware if they have a genetic disease they might pass on to a child. But I think in the past we used to say there's nothing we can do about it, and it's clear now there are lots of There are a lot of things going on that, done right, funded right, done openly, could lead to a set of tools that could create that kind of a golden age. So, in Cambridge, we have such incredible intellectual capital over here, all the biotech companies. And it seems as though people and data are siloed off from each other. And I was wondering what thoughts you had —you had on connecting people together so that if somebody is looking at some data set and another person is looking at related data sets, someone experiences— Well, that data platform I was referring to, that thing called Terra, is an example of that. It's meant to be able to host many, many, many people's data and allow there to be common programs run on it. We can't force people to share their data. We set a good example. By all the data we generate, we put out there. The Human Genome Project was organized, as I told you, around the idea of freely available data. We set good examples. We want to at least make it possible. Right now, people say, I can't share data with you. It's on my computer. It's not compatible with yours. Often, that's just a smokescreen for I don't want to share data with you. And so we'd like at least to clear that smoke by creating a platform where where if you want to share your data, you have to flip one bit to make it shareable. Then we can give people a hard time in saying you can't manage to flip the bit. So that's the kind of thing. It's not trivial. And I want to respect the fact that people may not want to share for different reasons. But you can set norms. The Human Genome Project was the first big project that set the norm that data should be freely shared. And the fact that we sequenced the human genome doesn't mean we should sit on it. We released the data freely, even to the private competing effort called Celera Genomics that was trying to beat us to the punch. Their final paper, published in parallel with our final paper, also used all of our data. And we were very proud because that was our principle. Not only did we have our paper, we could point to their paper in which our data was like like half of all the data in it.

That's okay.

Yeah.

Another question that's come in from the internet. It's really curious today. Yes. The internet wants to know about human cohorts and factoring in the genetics of underrepresented populations, the fact that a lot of the data that we have comes from largely European population. Exactly. So how are we addressing that, both the Broad and the larger genetics community?

Well, it's appalling that the vast majority, 80-plus percent of the cohort, of the samples that have been collected come from European-derived cohorts. It's appalling. It's perhaps not so surprising, because the medical systems in European-derived populations, United States have been some of the better-funded ones. They've taken it out of those medical systems. But that's no excuse for the fact that we have to attend to the fact that we have an entire diverse world. So in the United States, there are a number of dedicated collections that the NIH did that sample particularly African-American populations. When what was called the Haplotype Map Project started, it was very important that it didn't just do European populations. It did the Yoruba populations in West Africa, which were very important sources of the African-American populations in the United States. It did Han Chinese and Japanese. It still is the case that these disease populations are being collected at a much faster rate in European populations, so that it's not catching up. We need much more serious, dedicated efforts to build population cohorts in many other countries.. And part of this International Common Disease Alliance, one of its commitments is to global diversity in that. And it has representatives on its organizing committee from Africa, from Asia. And none of this is trivial, because getting the resources to do it is hard. But it starts with acknowledging that it's very important. It will be unacceptable to have polygenic risk scores that will predict who's at high risk for heart disease, provided that you're of European descent.

Right? That's like not OK. I think we have time to take one more question from the room.

Yes, over there. Do you believe there will be a genome-enhanced human in future? Will there be a genome-enhanced human in future? I got to say, this idea of CRISPR editing babies to improve, in air quotes, humans, I'm pretty dubious about. As proud as I am about the fact that we read the human genome about 15 years ago, that's like saying, oh, I'm in like 7th grade and I've read Shakespeare. I can do better. We don't know. I mean, I'm just deeply dubious. Deeply dubious about efforts now to try to, quote, improve human beings by editing their genomes. I think we are deeply ignorant about those effects. It is possible to do it. We saw an example in Hong Kong of a scientist, He Jiankui, who actually did genome editing of two twin baby girls. It was an appalling, unethical experiment. Experiment in all its dimensions, from the consents that weren't obtained properly to the choice of what kind of editing that should be done. I'm not saying it shouldn't ever be done, but I'm saying I'm pretty dubious that most of the ideas that people have for, quote, improving humans are pretty woolly and poorly thought out. And if we're going to do any of that, it has to be pretty narrowly focused. Now, I happen to be on a commission that was appointed by the U.S. National Academy of Sciences, National Academy of Medicine, UK's Royal Society that's going to be meeting over the course of the next year to think about are there translational paths forward for some such things. And there may be some. My own guess is they ought to start pretty darn limited because there is a tremendous risk of hubris in thinking that we can just willy-nilly improve humans. I think we will cause damage almost All genetics starts with these simplistic views, and then we come to, over time, recognize these things are complicated and they need to be done with respect. Whether 3 generations from now this is happening, I think that's going to be up to the younger people in the audience. For right now, I'd like to make sure we don't go forward with the kind of mistakes one's seen in the past. And one approaches this in a cautious, careful, limited way. Anyway, that's Broda 15. There'll be a bunch more Broda 15 lectures. Come back for them, and then put on your calendar Broda 20, Broda 25, and other things. Thank you very much.