SV4GenDis

Structural variants (SVs) are genomic variations of 50 nucleotides or more. They can take multiple forms, from simple deletions and insertions to complex rearrangements. Despite their known link to rare genetic disorders and some common diseases, they remain absent from most large-scale genomic studies of diseases, primarily due to the technical challenges in identifying them. Because they are typically not included in those genomic studies, it means that we have been studying only half of the variable genome, and it could explain why a large part of the genetic factors of diseases remains elusive. This transformative project addresses this critical gap by devising new methods to integrate SVs into genome-wide association studies. We will first harness the power of short-read sequencing data and cutting-edge pangenomic tools to genotype SVs and conduct a genome-wide association study within the expansive dataset of the UK Biobank. The UK Biobank gathered a prospective cohort of half a million people in the United Kingdom, sequenced their genomes, and is sharing this data and clinical information with the research community. We will first analyze SVs in 11,000 cases with inflammatory bowel disease and as many controls to boost the interpretation of the many uncharacterized loci associated with the disease. Novel associations will also be identified thanks to our high-resolution study across a large cohort. Indeed, this project will mark the first comprehensive analysis of SVs within a sizable cohort for inflammatory bowel disease. Then, we will develop novel SV imputation strategies to predict SV genotypes based on neighboring small variants. Imputation uses a catalog of known haplotypes to predict the genotype of variants from low-resolution assays like SNP-arrays or low-coverage sequencing data. Structural variants are not currently well imputed because we are lacking large high-quality haplotype catalogs and specialized methods to model their complexity. However, SV imputation has tremendous potential because it would help integrate them into the many existing studies that used these lower-cost assays. Here, we will leverage a comprehensive haplotype catalog that we are currently building in the Human Pangenome Reference Consortium and the haplotypes across the more than twenty thousand individuals in Aim 1. This unique haplotype catalog will be combined with new innovative methods that specifically model SVs. Briefly, we will extend existing approaches to handle more than two alleles and to take into account dependencies when multiple variants overlap. A particularly promising approach will build on our recent work with pangenomes to naturally integrate all types of variants and haplotypes. In addition to imputing SVs from single-nucleotide variants, we will extend our approach to low-coverage short-read sequencing to enable cost-efficient analysis of large cohorts. After a comprehensive evaluation and benchmarking, we will impute a large set of SVs across all 500,000 individuals of the UK Biobank with our cost-efficient imputation strategy. The SVs, observed for the first time across such a large cohort, will be tested against the rich phenotypic information in this biobank. For this third aim, we have selected twelve common traits because of their medical relevance and known missing heritability. The goal is to maximize the added value of integrating SVs by focusing on phenotypes with a strong genetic component and clinical importance. These traits include, for example, hypertension, asthma, type 2 diabetes, and heart disease. We will showcase the power of our new approach on twelve large genome-wide association studies for diverse traits and diseases. In addition to these applications, we will share our imputed variants with the other researchers working on UK Biobank, so that all traits and diseases benefit from a more complete genomic picture of common variants, now including most SVs.