 All right, guys, I think we're going to get started now. We have a number of medical school medical student presentations. The first one is going to be by Michael Simmons, who's a fourth year medical student from the University of Arizona. And I'm just going to talk about a unique aspect of the ARED's tribe. Thank you very much. As Renee said, my name is Michael Simmons. I'm a fourth year medical student from the University of Arizona. And the topic of my presentation today is investigating deep genotype-phenotype relationships in age-related macular generation. And for this, we use data from the age-related eye disease study trials. So let's see. Actually, I'm going to need to pull out a keyboard for one second. OK, my mentor in this research is Dr. Emily Chu, who's the deputy director of the Clinical Trials Branch at the National Eye Institute. My co-mentor is Dr. Zhang Lu, who manages the text-mining division at the National Center for Biotechnology Information. He and his team are responsible for the search functionality of PubMed. My research was made possible through the sponsors of the Medical Research Scholars Program at the NIH, which is a year-long residential research internship that I recently completed at the NIH. So of course, we're talking about age-related macular degeneration today. And this is the leading cause of blindness in the elderly. It's a chronic neurodegenerative condition that results in the progressive loss of the retinal pigment epithelium and loss of photoreceptors. Although there's no definitive treatment, we have made a lot of progress recently with anti-vegeta agents. And the age-related eye disease study trial and its follow-up, AREDS II, showed that the use of high-dose antioxidants and zinc delayed the progression to advanced stages of the disease. There's a very strong genetic component of AMD, and this is going to be the topic of our discussion today. So recently in 2016, the largest genome-wide association study of age-related macular degeneration was completed. In this study, Fritz et al. took 16,000 people with AMD as cases and compared them to 18,000 controls without AMD. And they looked at genes by with broad gene sequencing that segregated with the presence of AMD. You can see here in this Manhattan plot, so named because it looks kind of like a Manhattan skyline, that the x-axis contains various single-nucleotide polymorphisms arranged by chromosome. And the y-axis has p-values of the strength of association between each of those SNPs and the development of AMD. This study identified 52 individual SNPs or genetic loci that associated with AMD. And 34 of those were most representative, including complement factor H, which was the first gene ever to be identified in relation to AMD, arms 2, and it's accompanying gene at this locus, HTRA1, and the MMP9 gene, which is the first gene in a genome-wide association study to be found to associate with a subtype of AMD, that of wet macular degeneration. This finding here is what interested me most. It seems startling to me that it took the largest study ever to find any significant association with a gene and any of the subtypes. Because AMD is a very heterozygous condition, it seemed that we might have found other subtypes associated with specific genes before this. So we asked the question, are there differences in how these known genes discovered by the Fritz et al study? Are there differences in how these genes predispose to age-related macular degeneration? To answer this question, we developed a modification of the genome-wide association study construct, which we termed a deep phenotype association study. In this, we took a collection of people, in this case, participants in the AREDS II trial, and we split them into groups of cases and controls, not by the presence or absence of AMD or any other condition, but rather by the presence or absence of a specific allele. We then looked at a broad array of phenotypes found in the deep information in the AREDS study and looked at how one gene associated with many different phenotypes. This is opposite of what a genome-wide association study is, where they look at many genes in association with one phenotype. And we generated from this the familiar Manhattan plot. Although, I have to say Manhattan seems a little presumptuous for this tiny raffin comparison to the Fritz et al study, so it might be more appropriate to call it a Pocatello plot or something like that. So, what are the deep phenotypes that we looked at? Well, we took phenotypes that were measured by the central imaging center for the AREDS trials. And these phenotypes included things such as, or could be clumped into categories such as drusen with soft drusen, calcified drusen, reticular drusen and drusenoid pigment epithelial detachment as examples of these deep phenotypes. Autofluorescence abnormalities with irregular background autofluorescence, decreased central autofluorescence, geographic atrophy, including central geographic atrophy as a distinct phenotype. Pigment abnormalities, including hyperpigmentation or depigmentation, and the abnormalities associated with new blood vessel growth, such as laser treatment scars, hemorrhage characteristic of AMD, hard exudates, or visible vessels. We took each of these phenotypes and then conducted a pairwise chi-squared analysis between that phenotype and each of the known genes that were discovered in the Fritz et al study, the 34 independent genetic loci. Because chi-squared analyses need to take place on a one-to-one basis, or basically, they need to have binary classification of variables, we made the assumption that if a patient ever received a read of a specific phenotype, then they had that phenotype for sure. We also assumed an autosomal dominant model, meaning that if the patient possessed even one allele of the independently associated loci from the Fritz study, that would be enough to manifest in a phenotype in the patient. So from this, before I go into the data, I'd like to give you the baseline characteristics for we've gone for this study. So we took patients from the ARES II trial who had genetic testing, which amounted to 1,826 patients. Of these patients, about 1,000 were female. Nearly all were white, 792 had never smoked, and a little more than half had had some college education. This table here shows you that the various treatment arms of the ARES trial were represented equally amidst those patients who'd had genetic testing. And this last table shows you that the mean age of patients in our study was 73 years, and the average follow-up is about five years. This is the raw data from our deep phenotype association study. On the column, or the rows of this table represent each of the 30 deep phenotypes that I discussed earlier, and the columns are each of the 34 independently associated genetic loci. The table is a heat map with the various colors showing you the relative values or relative p-values of the strength of the association. Green cells had p-values that were greater than .05, and cells that were less than .05 were color coded from red to yellow, with yellow being the most significantly associated. For the rest of my presentation, we're gonna focus on just one gene and the various phenotypes associated with it, and this is the ARMS2-HTRA1 genetic locus, and the phenotype that was most significant in this locus was hemorrhage characteristic of AMD. I'm gonna show you the Pocatello plot for this deep phenotype association study, and you can see here the red line across, first of all, x-axis here represents each of the 30 deep phenotypes arranged in cluster by the different classification. The right side cluster, the first cluster is out of neo-vascular phenotypes, so you can see that there seems to be an association of neo-vascular phenotypes of the presence of the ARMS2-HTRA1 genetic locus. The y-axis is still inverse log p-values of the strength of association. The red bar across the top represents a Bonferroni correction threshold to account for the nearly 1,000 statistical tests that we performed and the increased likelihood of getting a false positive. Basically, we increased the threshold by which we would consider anything significant to make sure that we wouldn't consider too many things significant from performing multiple testing. The thing, only one phenotype passed our rigorous correction threshold, and that is hemorrhage characteristic of AMD, and this is the table from which we derived our chi-squared analysis and odds ratio. You can see here that there were 1,826 patients, just like we showed earlier, and of those, 423 developed hemorrhage characteristic of AMD. About 75% of those patients had the ARMS2 gene, and only about 25% did not. And from this data, you can see that there's an odds ratio with patients with the ARMS2 gene, about 80% more likely to have hemorrhage with AMD than those who did not. And the PVA for this was 1.17 times 10 to the minus sixth. Now, there's a known association between the ARMS2 genetic locus and carotoneovascularization, and it's common knowledge that carotoneovascular processes are more likely to hemorrhage than normal blood vessels. So we wondered if this association was simply related to the connection between ARMS2 and carotoneovascularization. To see if this was the case, we took all the patients from those with genetic testing who had carotoneovascularization and repeated the test in only that subset of people with CNV. We would expect there would be no association if ARMS2 did not predispose to hemorrhage because all these patients had CNV, so they should be equally likely to hemorrhage. What we found, though, was in fact that the odds ratio for hemorrhage, even in a population of people with CNV already, was the same. People were still about 75% more likely to hemorrhage if they had the ARMS2 genetic locus, which means that this increased likelihood of hemorrhage is independent of CNV status. And the p-value for this was still significant. So the next thing we did was we wanted to look at this association in a new subset of patients. Since this is a secondary analysis, it's likely that we might have beginner's luck to find a spurious association. But the AREDS trial, which preceded the ARDS2 trial, was a completely separate group of patients. From this trial, there were a few that were included in the AREDS2 trial and we excluded those patients. We also excluded all people who did not meet the inclusion criteria for the AREDS2 trial. So they had to have large drisnet baseline, and then we took of those only people who had CNV already. This is the table for that, and you can see that there were 748 people who met our inclusion criteria. 509 of these had hemorrhage, which is a bit higher percentage than the last one, and that makes sense because back in the 1990s when ARDS was conducted, we didn't have anti-vegeth agents, and so there was more hemorrhage in general than there is now. Of this 509, nearly 400 patients had hemorrhage with the ARMS2 variant. And the odds ratio was again consistent in this new subset of people with CNV. P value was still significant. And so from this we concluded that first of all, individual genetic variants do seem to exert specific deep effects on the development of AMD. Second of all, we found a very strong relationship between the ARMS2 HTRA1 locus and the development of hemorrhage characteristic of AMD. These are my acknowledgements, and I'd like to take any questions. Thank you. Dr. Brinsting. So can you talk about the challenges of the fact that AMD is a disease in evolution? So you're picking phenotypes that you know people who have bruising are eventually going to advance to the other disease. Does that, to more advanced diseases? Is that the problem or not? Actually, I think that this is one of the strengths, I think this is one of the strengths of our study and of the genome-wide associations that get concentrated in general. The connection between genes and phenotypes is kind of a black box. We know that genes are the building block for everything else in nature, so they've got to have some connection with disease. And that's what the genome-wide associations they enter really well. It says, we don't need to know anything in the middle. We can just say, is there an association between the presence of this gene and the presence of this phenotype? And that's what we've continued to do with this deep phenotype of the species. We don't know exactly how, for example, ARMS2 leads to hemorrhage characteristics of AMD, but without knowing any of the intermediates, we can say that there's a strong association between the presence of the gene and the presence of this fine phenotype, which is an advanced phenotype. I think that one of the reasons why we're able to find significant findings in it at any rate is because of the size and quality and duration of the agent. So I was without that, with just one-year study or a small subset of patients, we probably wouldn't be able to detect it because as a disease of aging, it would take a very long time for us to see the chronology of progression of phenotypes in AMD. Any other questions? Thank you very much.