 Without further ado, let's begin the session. My name is Mark Ramis, and I'm one of the Associate Scientific Directors for the Canadian Longitudinal Study on Aging. I'm also the leader of the webinar series along with our communications officer, Sue Johnston, and I'm an associate professor in the School of Public Health and Health Systems at the University of Waterloo. Today, it brings me great pleasure to introduce Dr. Brent Richards, who's going to talk about the genetics of osteoporosis, an aging-related disease. Dr. Richards is an endocrinologist and a genetic epidemiologist at the Jewish General Hospital, which is affiliated with McGill University in Montreal. He has participated in and sometimes led several recent global efforts to understand the genetic determinants of osteoporosis. Dr. Richards has published 94 peer-reviewed articles in many prominent journals, such as Nature, The Lancet, and Nature Genetics. He is funded by CIHR, the Public Health Agency of Canada, the FRSQ, which is the Quebec-based Provincial Healthcare Funding Agency, and several other organizations as well. He's received funding from GlaxoSmithKline and Eli Lilly on the pharmaceutical side, and he will again speak to us about genetics of osteoporosis. So just a couple of brief words on how the webinar is going to unfold. Dr. Richards will speak for approximately 40 to 45 minutes. After that, there will be time for questions from the audience. You can't ask your questions orally. We've disabled the talk feature. That is to prevent feedback when too many people have their mics open at once. So instead, please type any questions that you have in the chat feature. It's located at the bottom left-hand corner of your screen, and I will read out the questions to the whole audience, and Brent will be able to answer them. So I'll stop talking and now I'll introduce Brent. Thank you very much for agreeing to present this webinar today, and I'll turn it over to you now. Thank you very much, Mark. I hope people can hear me. Maybe somebody could just, maybe Sue could just type into the chat function and let me know that you can hear me. I am very happy to give this talk. I'm going to talk about the genetics work that we've done on osteoporosis, and the reason why I wanted to do this for CLSA is because we have secured funding for genome-wide genotyping of 10,000 people in the CLSA, which is a very major investment in the largest genome-wide genotyping resource in Canada and amongst the largest in the world. So I'm very excited about possibilities of genetics in CLSA. We are also working to secure funding to genome-wide genotype the remaining 30,000 people, excuse me, the remaining 20,000 people, bringing the total sample size up to 30,000 people with DNA. So I'm very excited about the possibilities of genetics within CLSA. I'm going to share with you some of our experience in the field, and I'm going to apologize because I have a bit of a cold, so if I have to stop and take a cough, please excuse me. Okay, so just to get things started, I don't have any disclosures, but I just wanted to talk about the outline of what I'm going to talk about for the next 40 minutes or so. I'm assuming that at least one or two people in the audience don't do genetics on a day-to-day basis, so I'm going to describe some broad strokes as to why genetics would be helpful. Why does a genome-wide association study, or as we call them, a GWAS, and then I'm going to finish with some work that we've been doing recently on whole genome sequencing. I should also mention that we are trying to get a portion of CLSA whole genome sequence to enable some of the studies, such as the one that I'll describe at the end of the talk. So if we think about disease or why my patients get sick or why you or your parents get sick, it really can only be due to three factors. So one is genetic effects, the next is things that we are exposed to in our environment, and the third thing is the interaction of these two. Other than divine intervention, there's really nothing else that I can think of that causes disease. And so if you're interested in treating disease, I think it's good helpful to understand its cause, and if you can understand its cause, then sometimes you can help people. When we talk about the difference of the proportion of risk in a population that is due to environmental versus genetic effects, what we do is a heritability study. And in this slide, I'm introducing a concept called BMD, and that's bone mineral density, and that is simply a measure of the density of one's bones, and it is the most clinically relevant predictor of osteoporotic fracture. Now, BMD is very heritable, so it has a heritability estimate of up to 80%. And osteoporotic fractures are less heritable, but still importantly heritable, with a heritability estimates of around 50%. Now, these are not precise numbers, different studies report different numbers. The point that I'm trying to make is that genetic factors play a very important role in predetermining our bone density and our risk of osteoporotic fracture. What I next wanted to talk about was really, so if genes are important, well, why would we bother studying them? Well, I think one of the reasons why genetics in humans is important to me is because it isn't humans. As you'll see, I spent a lot of time on mass models. I believe strongly in their relevance in some contexts to human disease, but I think as a jumping off point, if we can start with human data and then move to a mass model, it can be much more relevant than using murine data and moving to a human model, and why is that? Well, because we know that things that are relevant to human disease should probably best be found in humans prior to investigating their function within murine models. The other thing that I find useful about genetics is that it's agnostic. What does that mean? Well, it means that in modern genetics we approach and experiment without any preconceptions about what might be true. We do this because this enables us to find things that we didn't know that we were looking for, which can lead to exciting insights into disease etiology. Sometimes in human genetics we can identify things that are causal, and they're very rarely confounded. So by confounding, I mean that the genetic factor is associated with an outcome only through a marker which is not in the causal pathway. The last thing why I think human genetics is important or sometimes helpful to look at is that it is not influenced by a reverse causation. By this I mean that the disease itself does not influence your genotype, and that is often not the case in observational epidemiology. So very rapidly we can identify what are potentially causal pathways in disease etiology in humans using human genetics. So human genetics asks a very, very simple question. What it generally tends to ask is what are the genetic variants that are associated with the variation in human disease risk or human trait? And the way that we do such a study is we simply just do an association study. If many of you are involved in epidemiology or in murine models of disease, these are quite familiar with you. So you observe some variation in a population, in this case genotypes, and you ask if that's associated with variation in disease outcome. What has been quite exciting in human genetics is that there's been an emergence of new tools, and those new tools have enabled us to ask this very simple question on a very large scale. And those tools are very similar to this tool that's shown here in concepts. So this is Galileo, and this is a telescope that he made in the 1600s, and he used this telescope that he made with his own hands to look up into the night sky. And when he looked up into the night sky, what he saw was Jupiter and its moons, and what he could see as he looked over the evening that the moons of Jupiter seem to be moving around Jupiter itself, which of course changed our notions of the universe that smaller heavenly bodies tend to revolve around larger heavenly bodies and changed our view, or changed humanity's view of a geocentric universe to a heliocentric solar system, which really was probably many people looking into the night sky with a telescope could have seen that. But the real advance for Galileo was that he just simply had a better telescope. And if we move forward 400 or so years, this is a view from the Hubble telescope. And so just through advances in technology, we're able to capture a much more queer view of the universe. And so the exactly the same thing happened, has happened in human genetics, where simply advances in technologies has allowed us to do, allowed us to look not everywhere at once, but many places at once to try to understand the genetic determinants of disease. So what do we need to do these sorts of studies? Well, we measure genetic variation everywhere at once. And then we simply measure a phenotype that is at least partially heritable. So if you're interested in your cognitive disease, then all of you, excuse me, all you do is you measure a phenotype that's relevant to your disease. And then you associate those two different types of variation with each other. So how do we look everywhere at once? Well, the first thing that we do is we genotype somebody on one of these gene chips which are shown on the right-hand side. And we do this by genotyping a million SNPs across the entire genome at one time. And in many improved versions of SNPs, we can genome-wide genotypes, several million polymorphisms at one time. And what is a SNP? Well, that's just a single nucleotide polymorphism, which is just simply a base pair change that can be measured in the human genome. So one of the things that's important to understand as we move through descriptions of these studies is that through the study designated as a genome-wide association study, what we're measuring here is common genetic variation. What do I mean by common genetic variation? I mean that genetic variation, which is present in at least 5% of the population, or has the minor allele of that genetic variant is present in about 5% of that population. And we know that if we're looking at common variation in the population, we must have some certain expectations of what we'll find. And those are really just driven by natural selection. So this is just to remind you about natural selection. If you take a population of different cells that have different levels of resistance to antibiotics, as shown by their different colors here, and you expose them to the antibiotics, well, then the next generation of cells will be those that are highly resistant to the antibiotic. So what you're doing is you're applying a pressure to a population and changing its allele frequencies. Now, we know that if natural variation had introduced into the population something that has a large effect on disease risk, that that would not be allowed to become common in the population. And so we know that if we're going to look at common genetic variation in the population, that we would expect to find small effects. So with that as the beginning, I'm going to talk about what a genome-wide association study is and how one does them. So I'm sorry, my telephone keeps ringing here. Excuse me for just a moment. I'm just sharing a webinar. Yes, I can't talk. I'm just sharing a webinar. Thank you. So to do a genome-sorry, but to do a genome-wide association study, all you do is you take a population such as the people in this picture who are watching a tennis game at Wimbledon. You select from that population individuals who may have a disease or may not or may be controls for that disease, or you measure something in them such as their height. And then you take DNA from those individuals. You put it on such a gene chip. These chips allow you to measure genetic variation at millions of different single nucleotide polymers at once. And here's an example of such a single nucleotide polymers. So in some people we have a CG base pair and some people we have a TA base pair. And so this means that this base pair is polymorphic. And because there's variation in the population, we can simply associate that variation with variation in height or case control status. And we just very simply ask the question, do the people who have the CG, do they tend to be taller than the people who have the TA? And that in its simplest terms is what a genome-wide association study is. Now, there's lots of problems that come from looking at so much data at once. I'm not going to get into all of these problems, but I'm just going to highlight some of the more important problems. The first is that we're really drinking from a fire hose. So what we're doing here is we're assessing, let's just say, a million single nucleotide polymorphisms. And so if we use traditional association test statistics, applying a P value of 0.05 to our data, that would mean just by chance that 5% of 1 million SNPs would be associated with our phenotype. So we would then say, let's say, write a paper to nature and say, wow, we found 50,000 single nucleotide polymorphisms that are associated with our outcome, which is of course not true. These are just all due to chance. And so what are we going to do about this from? How are we going to deal with the fact that we have a million single nucleotide polymorphisms and we know that we cannot accept a P value of 0.05 to be reasonable evidence for strong association. So what we do in the first instance is we make those P value thresholds more stringent. In fact, we drop the P value threshold down to approximately 5 times 10 to the minus 8 and dropping it down to that level accounts for the number of multiple statistical tests that we've done. And the next thing that we do is we do replication. And we go to another cohort around the world and we replicate our findings. Or we look for the same association within different cohorts around the world and see if we find the same thing. And if we do find the same thing, well then that's evidence that what we have observed in the original study is unlikely to be a false positive. So this creates a lot of problems. This creates a lot of opportunities. And really what this creates and I think will create a lot for the CLSA is the importance of international collaboration. And so I bring up this slide just as a joke because what you can see is that when we all get together in the bottom of some hotel somewhere around the world and talk about the genetics of a disease of interest, what we're also bringing to the table is all of our prejudices about what people who used to be our collaborators are like. And it's when we overcome those prejudices and start to work together as a team and start to harmonize our data collection, our methodologic strategies. We find that not only are we able to replicate each other's findings, but we're also able to provide much more value to the scientific literature. So I'm going to show you now what happens when we do such a study and we increase our sample size. And I'm going to do that in the context of looking at lumbar spine, bone marrow density, that's just looking at your spine and its bone density. And this is called a Manhattan plot. And what I'm going to show you is what happens as we increase our sample size. So that's the end here. And what each of these dots here represents is a negative log 10p value of an association of a single SNP with bone marrow density. And suffice it to say that as this dot moves further and further up this y-axis, that shows stronger evidence of statistical association. What we have laid out in the x-axis is genomic position. So this is the first chromosome, the second chromosome, the third chromosome. So things that are closer together on the x-axis are closer together in geographic space in the genome. And what we're showing here in green is what we will declare as genome-wide significance, as I mentioned, is a p-value of less than 5 times 10 to the minus 8, which accounts for all of the independent statistical tests that we've done in our genome-wide association study. So you can see that nothing achieves significance here using just 5,000 people. So let's increase our sample size to 8,200. What's kind of exciting? Here we have something that's new. We'll keep going to 13,000. You can see that now we have lots of these little spikes coming up from the data. And we get to 16,000. We have a real bubbling up of data. So we have really strong associations at several different loci in the genome. And when we get to 19,000, we have what we call a Manhattan plot. And a Manhattan plot is just supposed to look like the Manhattan skyline is that we have some baseline level of noise. And then across that skyline, we have these huge skyscrapers, which are screaming out to us that there is some sort of... So there's something going on at the genome underneath this peak, which is strongly influencing lumbar spine with bone marrow density. So we've done studies now that incorporate up to half a million individuals. And I'll go through some of these, but you can see that as we increase our sample size, because of the small effect of the genetic variance, we start to see larger and larger effects. And so this has been done not only in osteoporosis, but in other diseases as well. In genome-wide association studies have really been... They've really become the backbone or the workhorse of modern human genetics. And so just to give you sort of an idea as to how that looks on the ground, if we go back to 2005, this is... Sorry, I'm just going to show you what happens over time, where this is the human genome, and each of these is a chromosome. And I'm just going to put a little dot on each of these chromosomes where there has been an association with a different trait at a genome-wide significant level. So if we go back to 2005, we have only two loci that were reproducibly associated with any common disease. And this really was prior to the genome-wide association era. And we can see that despite really over a decade of investigation and investment, there was almost nothing that had been reproducibly associated with any phenotype. And watch what happens as we move into the genome-wide association study era. And each time I push this button to move the slide forward, we're going to move forward three months in time. So you can see that the first genome-wide association studies start coming out, and then more and more and more until we move through time and end up at 2014. And you can see that we have really been able to map a lot of the human genome to different parts, different risks of common disease or traits of relevance to common disease. So what does all of this mean? Well, I'm going to bring up some things that I think are important to understand about genome-wide association studies, both in what they're able to do, but also importantly in what they're not able to do. So one of the promises of the Human Genome Project was to try to identify regions of the genome which could lead to new drug targets. And so what we did was, in osteoporosis, we already have some drugs that we use to treat in the clinic. And so if our agnostic approach, which doesn't assume anything in the human genome, is associated with osteoporosis, we are just going to ask the very simple question. Do we identify through a genome-wide association study the gene which is the target of the drug which we already know is important for bone mineral density and osteoporotic fracture? So we did this here. So there's all these different classes of drugs, and they all have different targets. Some of the targets overlap, such as selective estrogen receptor modulators or estrogen, but some of them are unique. And so now we just ask, through our agnostic approach, do we find these targets? And if so, then that would provide strong evidence that the novel loci which we have identified that are not at these known osteoporosis-relevant drug targets, but those novel loci would contain some proteins which would then be relevant for drug development. And so when we do this, we find that the drug targets are identified for most of the clinically relevant drug targets for osteoporosis. We do miss pharnosolpyrophosphate for bisphosphonate, and of course we cannot identify strontium because it doesn't have a drug target, because of course it's just an element. And so is this true across diseases or particular to osteoporosis? Well, this is work that was done by GSK, which was published this year in Nature Genetics. And what they did here is they did a very similar approach across different diseases. And what they demonstrated was that if you look at the drug target, if a target has support from Human Genetic Genomic Association Studies and the online Mendelian inheritance of man or rare disease genetics database, so if a drug has support from GWAS or Mendelian genetics, does it increase the odds that that drug would have become a commercial success, or did it achieve regulatory approval either in the European Medicines Authority or in the FDA? And you can see that if you do have genetic support for a drug target, you have a much higher odds of successful drug development than if you do not have support from Human Genetic, suggesting that Human Genetics is a very useful source of information in which to base your decisions upon what to target to treat the disease. So with that, sorry, then the next thing that I wanted to bring up was one of the limitations of Human Genetics. So many people wrote many PhD theses talking about the ethics of the Human Genome Project and how this will influence our ability to identify people who are going to get a disease. So we did this within osteoporosis, and this is a receiver operator curve, and so I'm a clinician and I get quite excited if the receiver operator curve looks like this and covers a lot of the area of this graph. If there is no utility of a diagnostic test, then the graph would lie along the line of unity here. And so you can see there's a very big difference between black and red, and I got very excited about that because I thought, okay, this is early days, but we're at least starting to get to this part of the square where we can actually use genetic information to predict who is going to get osteoporosis. And then I looked at the legend here. And what you can see is that my fancy genetic score using all the best genetic information from across the world barely beats chance alone. Whereas if I just use age and weight to try and guess who's going to get osteoporosis, I do a much better job than the genetic score. And I can tell you if I combine the two, that I just sit barely above this line of age and weight. And so we see this not only in osteoporosis but across most diseases that we cannot predict using common genetic variation who will get a disease. So it's very safe for you to be able to tell your patients or tell patients that are subjects that are in your studies that this information very, very rarely is able to help us understand who will get a disease. So that's one of the limitations of these sorts of studies to date. Okay, so with that as an introduction of what a genome-wide association study is, how they work, and some of their benefits, I'm now going to talk about whole genome sequencing. So whole genome sequencing is different from genome-wide genome typing. In a GWAS, we would look at approximately, as I mentioned, a million or seven million or using different strategies up to 20 million genetic variants. In whole genome sequencing, what we do is we actually capture approximately three billion pieces of information per person. So a lot of these base pairs that we measure are not variable. So they're conserved between people, but some of them are variable. So we use that variation within base pairs that, again, we try to associate with variation in disease status. So that's a bit of a, well, it's a bit of a programming task. It's a lot of work has to go into trying to store this data. So why would we actually try to find or use whole genome sequencing to identify other genetic variants that are associated with disease? One of the rationale is here, and that has to do with the allelic architecture of a disease. So within a genome-wide association study, what we have done is we have looked at genetic variants that have a minor allele frequency down to about 0.5% or, excuse me, 5% or 10%. And we do know that, as I mentioned, most of these things have very small risk of disease. And that's because they have survived natural selection and been allowed to become common within the population. As we move down to the lower end of the allele frequency spectrum, it has been hypothesized that what we'll see are things like this guy, where it has become rare, but has a much larger effect. Or this guy, which has a very large effect upon disease, and maybe the reason why it is rare is because natural selection has made it rare because it has a strong and deleterious effect upon disease etiology. So with that as a rationale, I recently led a program where we published this year a whole genome sequencing study where we identified E1 as a determinant of bone density and fracture. This involves collaboration in many parts of the world. As you can see here, predominantly Europe, North America, and Australia. And we call ourselves GFOS, which is the genetic factors for osteoporosis. And through this study, we tried to tackle that low end of the minor allele frequency spectrum to try to identify large effects upon osteoporosis. So here is our study design. So what we did was we took 2,800 people who had been a whole genome sequence through a program that I have co-chaired called the UK 10K program. We also undertook whole exome sequencing of 3,500 people. Whole exome sequencing is different from whole genome sequencing because in whole exome sequencing, what you do is you just identify base pair changes that lead to differences in coding amino acids, whereas whole genome sequencing measures everything in the genome. And then what we did was we took people who had been genome-wide genotyping and through a process called imputation, we used what are called haplotype blocks, which I won't describe in detail, but we'd be happy to answer any questions upon. We used haplotype blocks to try to impute missing genotypes, and particularly those that were uncommon or rare. We did that in 26,000 people. We pooled all the data together and undertook a single variant and something called SCAT tests. SCAT tests I won't talk about, but we did this in 33,000 individuals approximately. And then we sought replication for our finding in additional 16,000 people. We then tested the association of our lead variants in up to half a million individuals, and we undertook functional validation using mouse and rat models as well as intelical analysis from different consortia that have tried to understand the impact of functional non-coding genetic variation upon disease etiology. And I'm just going to show you one of the stories that we were able to tell from this data. There's many different stories to tell, but we don't have much time, so I'm just going to talk about one of the findings that we thought was quite interesting. So we identified, I'm not sure if you can see this bar here, but maybe I can move this. We identified a single nucleotide polymorphism that has a T allele that has a pretty competent association with bone mineral density and lumbar spine. What was interesting to us is that the effect of this allele is measured in standard deviations. So this one single allele moved bone mineral density by a quarter of a standard deviation, which is a huge effect, much larger than what we would expect for low frequency, or excuse me, common genetic variation, and much larger than most of the environmental determinants of us or process. And so if you had two of these, your bone mineral density would be predicted to be increased by half a standard deviation. So we thought that was exciting, but being skeptical, we really wanted to have a replication of this finding before we saw a high level publication. So we replicated this in 15,000 individuals. We found that the effect size was moderately decreased, but we found again strong evidence of association at this single nucleotide polymorphism. And when we combined this data, we found that the beta was particularly very strong with a very strong evidence of statistical association across 40,000 people. And this allele had a minor allele frequency of 1.6% in the general population. So that's one of the stories that we could tell, but also when we looked at other genetic variations, we found equally impressive statistical findings, but also some very large effect sizes as well, but not as large as what we had identified for this variant for this single nucleotide polymorphism here. So in blue here, what we've done is we plotted the effect size in standard deviations of all of the genotypes or all of the steps that we had previously identified for GWAS. And on the x-axis, we've plotted their allele frequency. So you can see, as expected, the single nucleotide polymorphisms that have a high allele frequency tend to have small effects, and their mean here is in blue. As we move down to red, which are the novel low side that we've identified, we can see much larger effect sizes such that the top set that we have identified has an effect that's fourfold larger than the mean of what we've identified previously and twofold larger than anything that we've seen before. And so what we can see is that the data is starting to recapitulate our hypothesis here, that as we move down the minor allele frequency spectrum, we see larger risks of disease. So bone marrow density is interesting, but what my patients actually care about is fracture. So what we did was we looked in up to 500,000 individuals what the effect of these variants near this gene called EN1 was on fracture. And we could see that we had very strong evidence for our lead variant for fracture and it decreased risk of fracture by 15%. And you may recall that it was the T-wheel that increased bone marrow density and it also decreased the risk of fracture, which is of course what we would expect. And so this means that the people who are lucky enough to get two of these alleles decreased the risk of fracture by 30%, which is larger than really any drug that I use. The effect size is larger than any drug that I use in my clinic. And so we thought this was interesting. And the next sort of steps that we wanted to try to understand is, okay, we have these base pair variations, but what is the gene that's actually underlying this association? So what we did was we looked at the three-dimensional organization of the genome. So many people think about chromosomes as straight lines, and this is what's represented here, but they're not actually straight lines. They're actually like cooked balls of spaghetti that just flop all over each other. And that flopping allows different regions of the genome to touch each other, which otherwise in a linear two-dimensional context would not be touching each other. And so all of the single nucleotide polymorphisms are in red that achieve genome-wide significance. And you can see this is a heat map of this three-dimensional touching. They all live in the same area of the genome that frequently touches each other, and they center upon this gene called EN1. And we thought that was interesting because never before had EN1 been associated with a osteoporosis phenotype in humans or any bone phenotype in humans for that matter. However, there was some mouse evidence that had previously implicated EN1 in disease. So then we started asking questions about could it be EN1? So if it was EN1, we would want to know is it expressed at all in bone? Is it expressed in the right cells in bone? And what happens if we manipulate EN1? And so this is not work that I've done, but through collaboration with different functional teams around the world, we started to address these questions to have more confidence in understanding whether or not EN1 was the gene which isn't important in osteoporosis etiology. Our first clue came from osteoblasts, sampled in developing mice, and you can see that EN1 is expressed. And as we move across in days, there is a temporal change in EN1 expression. So that was an interesting and quick clue, but more interesting to me was that it was expressed in osteoblasts and not in osteoclasts. So you can see this here that we have clear expression across different days of development of osteoblasts with positive controls, but not at all in osteoclasts where it is present in a positive control in osteoclasts. Suggesting that EN1 is at least in one of the important cells in bone biology, but absent from another important cell in bone biology, osteoclasts. So now we asked, well, what about in tissue? So those are cells, but what happens in tissue? So we developed a fusion product to EN1, so you just take the promoter of EN1, you fuse it to a gene that causes a different color in histology, and in wherever EN1 is expressed, we see a blue staining. And we also co-stained with something called alkaline phosphatase, which is stained in red. And what alkaline phosphatase does is it tells us where osteoblasts are. And so as we look at the growth plate in a bone marrow, we can see that in blue, EN1 is well expressed. It's often but not always coincidentally expressed with alkaline phosphatase, suggesting that it is in osteoblasts within the growth plate in bone marrow. Interesting in the cortical bone, we see this in osteocytes. Okay, osteocytes are very difficult to grow in culture, but we can see that osteocytes have a strong, important effect upon... These are important sources for EN1 expression. And what are osteocytes? Well, osteocytes, if you think of the bone as a symphony, osteocytes are like the conductor. So they send out messages to all these other cells to try to influence the bone phenotypes. We also see this intravascular bone as well. And so we walked away from these experiments being rather confident that EN1 was expressed in bone. So then we thought, okay, if we really wanted to implicate this causally, what happens if we manipulate EN1? So we developed a model of EN1 that deletes EN1 in mice. So the problem is if you knock out EN1, you cause embryonic lethality. So what we did was we decreased the expression of EN1 through a construct that deleted itself almost perfectly whenever EN1 was expressed. And we spent a long time looking at the bones of mice that had much less EN1. So these are control slides. So these are in non-genetically manipulated mice. And these are... Here's a picture of the minerals. So this is the vertebral bone. This is the inside of the vertebral bone called the trabecula. And here's the outside, or called the cortex. And the trap staining tells us where osteoclasts are active. Now many of you may know that bone is not a static organism. Rather, bone is always turning over, always chewing itself up, and always making new bone. And the business end of the chewing up system is something called the osteoclast. And those osteoclasts are highlighted in yellow. And what happens now if we genetically decrease EN1 expression in mice? Here are these animals. And you can see that you can pretty much poke your finger through the vertebrae that lacks EN1. And if we look in the trap staining, we see much more intensity of yellow, suggesting that this loss of bone is due to the bone being turned over much more frequently by osteoclasts. So if we look at microCT, which is just another way to quantify this, we can see if we lose EN1, we lose a lot of the bone microarchitecture. And this resulted in increased scumetal fragility in the mice. So I'm going to try to wrap things up rather quickly now. But to summarize, genome-wide association studies generate well-replicated results which provide insights into the etiology of disease in humans. I want to point out that small effect sizes do not entail lack of utility. So the small effect sizes that you saw from our genome-wide association studies very precisely pinpointed agnostically the drug targets which we use in our clinics to treat osteoporosis, suggesting that small effect sizes are very important for drug development, which has been also replicated in work by Axel Smith Klein. I'm not going to go over this point because I didn't spend a lot of time on that in the talk. What are some...another part of the summary for whole genome sequencing? Well, whole genome sequencing can identify large effect size variants which does make functional dissection easier. And as well, as I've shown you, EN1 is a major determinant in a novel gene for osteoporosis and osteoporotic fractures. I'd just like to close with a few lessons that we've learned through this odyssey. One is that imputation is key and very important for being able to identify novel genetic variation associated with human phenotypes. And the whole genome sequencing program, the SCAT test or collapsing test were entirely unsuccessful. And importantly, whole genome sequencing is only necessary for a small proportion of the cohort, and that's exactly what we intend to do for CLSA. And lastly, that collaboration is absolutely necessary for these sorts of programs. Again, I'd like to emphasize that I think that the genome-wide genotyping that we're doing in CLSA now and 10,000 people will enable lots of these sorts of studies for many of the various phenotypes that have been collected in CLSA and those that might be of interest to you. And that whole genome sequencing, should it be funded, may be of further interest to the CLSA community and enable further scientific discovery. I'd like to thank my collaborators in the UK 10K cohorts team where we undertook whole genome sequencing, as well as my collaborators from the Genetics of Osteoporosis Consortium for our work on EN1. And I'd be happy to take any questions and hopefully we've left enough time to be able to respond to those. So thank you very much. Great. Thank you very much, Brent, for this very interesting and quite informative presentation. So now we will be happy to take questions from the audience. Again, if you have any questions for Brent, please type them in the chat feature at the bottom left of your screen and we'll read them out and Brent can answer. So I have one question and Brent, you just talked about it at the end of your presentation, linking some of this work to the CLSA. In CLSA you're going to be using a sample of 10,000 people and of course you'll be able to link the genotyping to all of the other data that are collected in CLSA. What might CLSA, because it's such a multidisciplinary study, how might that inform some of your work when you are able to look at the other CLSA data in addition to the genotyping data? How might that multidisciplinary perspective do you think inform some of the work that you guys are doing? Yeah, thanks, Mark. So one of the strong points of CLSA is that they have collected information across many different domains. Within those domains there have often not been done genetic association studies. So especially in many of the neurocognitive domains there have not been genome-wide association studies done. So I think that this provides an opportunity to CLSA researchers to take leadership around the world to undertake such a study in the very large sample that we've collected, which hopefully will be up to 30,000 people. And with a sample size such as CLSA provides, that will enable very confident inference of genetic variation which leads to variation in phenotypes. And then that information has been previously shown to be very relevant to the drug development process and when you can partner with people or do the functional dissection yourself. I think that that provides an exciting opportunity to identify entirely novel proteins and pathways which may be relevant for your disease of interest. So I think, Mark, the exciting part for Canadian researchers and people with access to the CLSA data is that the incredible breadth of data across the CLSA enables this to be done in a whole host of different phenotypes where it hasn't been done before. Great. Thanks. And earlier you touched upon ethics and it seems that the ethics related to some of this research tends to be very tricky. And what is the reason for that? Thanks for that question. So I think that a lot of the ethical issues that have been raised with genetics are in fact non-issues. Some of them are very important issues but I think that futuristic predictions made in the 1990s that a patient will come into my office and I'll be able to tell him with clarity at age 65 he's going to develop disease X, Y and Z is not at all supported by the data. So ethics boards and more educated ethics boards know that they don't need to worry about such ethical considerations. Other ethical considerations that are unique to genetics are frankly very few. One of the important things is of course data security. But of course that's important across all of our different CLSA phenotypes. What also is important is to ensure that no one can be de-identified using data from human genetics. But again, that is very important for all of the different CLSA phenotypes such as income, such as psychiatric history, etc. All of this is sensitive information but not anymore or less sensitive than genetic information. So I don't see any particular genetic ethical considerations that scare me within CLSA. In a lot of the ethical framework for doing such testing for conserving confidentiality of the study subjects as well as sharing this information with other collaborators around the world has already been worked at. Great, thanks. So we'll give the audience maybe another minute or so to ask any questions that they may have. Perhaps your presentation was just so complete that you answered everyone's questions. We'll just maybe give someone a minute or 30 seconds to see if they want to type anything. And while we're giving people an opportunity to type, I'll just take this chance to promote our next webinar. So it's going to be on November 16th from one to two in the afternoon eastern time. And Dr. Bill Leslie from Manitoba will be speaking on advances in fracture risk assessment. And he is also involved in the CLSA. So we do have a question from Alberta's Tomorrow Project. Have you used any obesity models with EN1 expression and weight change? No, so we have not at all used or looked at EN1 and its effect upon obesity. There's an important relationship, as people in the call may know, between obesity and osteoporosis. We have not looked at that at all. Maybe that EN1 is a mediator of the effect of weight gain on the skeleton. Although I can tell you that, at least in very gross observation of the mice, there's not apparent changes in body weight. But that would be something which would be of considerable interest to look at moving forward. Great, thanks. So I'm not sure if we'll get any more questions. Brent, thank you so much for presenting this webinar to us today. I'm not a genetics person, but I certainly learned a thing or two by listening and watching the slides. So it was very informative for me, and I'm sure for everyone else. So on behalf of all of us at CLSA, thanks again for presenting this webinar today. And we wish you all the best with your upcoming research involving the CLSA participants. Mark, thank you for your time. Thanks for making this opportunity. And I am always available by email if people would like to contact me for further questions. Fantastic. That's great. Thank you. Have a good day, everybody. Bye now. Bye.