Invited Talk by Oliver Stegle

Statistical methods to exploit high-dimensional phenotypes in human genetics

Abstract: The advent of genome-wide associations studies (GWAS) has already advanced our understanding of the genetic basis of many relevant traits, including human diseases. The traditional approach of GWAS analyses is predominantly based on simplistic one-by-one tests, correlating individual genetic variants with single quantitative or categorical traits. Thanks to technological advances and large international efforts, phenotypic profiles of individuals are increasingly deep and comprehensive, ranging from molecular to cellular traits to physiological phenotypes in people. These rich and high-dimensional phenotypes challenge established analysis methods and at the same time provide new opportunities for advanced and integrative statistical approaches.

In this talk, I will review these open needs in the field and discuss some of progress we have made in the area. I will describe advanced methods based on multivariate linear mixed model that enable joint tests between multiple variants and groups of phenotypes, while retaining scalability to cohorts with hundreds of thousands of individuals. These approaches both offer major gains in statistical power for detecting genetic effects, and yield improved interpretation of how genetic variability shapes phenotype correlations. In particular, I will revisit tests for genotype-environment interactions which we generalize to consider multiple causal variants and allelic heterogeneity.