Making Genomic Prediction of Complex Disease Equitable

Project Summary

Genome-Wide Association Studies (GWAS) are generated at unprecedented scale in order to link genetic variation to susceptibility to disease and other traits. But as the name suggests, GWAS only generate correlation data-and numerous challenges follow. Combining GWAS data with advanced population genetics tools provides a tremendous opportunity to learn about the genetics and evolution of complex human traits; but we need new statistical methods to realize the possible benefits.

My lab will develop such statistical methods and apply them to complex biomedical traits, drawing directly on my previous work and expertise in population genetics, statistical genetics, and computation. First, we will tackle a major concern hindering the adoption of polygenic scores-genetic predictors of complex traits, derived from GWAS: Their poorer performance in groups that differ-whether in genetic ancestry, environmental or social exposures-from the samples in which the GWAS was performed.

We will develop a mechanistic understanding of the determinants of the prediction accuracy of polygenic scores, thereby advancing complex trait genetics research for the benefit of all people-in particular historically underserved and underrepresented groups. Second, we will characterize gene-by-environment interactions in complex traits. There is ample evidence that such interactions are common but evidence from GWAS has been underwhelming.

We propose a new approach for characterizing gene-by-environment interactions, that might solve this apparent discrepancy: A model that expects concerted changes in magnitude of effects across a large set of variants, e.g. in response to an environmental cue. Where this model fits, it will be germane to phenotypic prediction and to the problem of polygenic score portability across groups.

Finally, we will leverage the results of the research described above to study natural selection on complex traits in recent human history. We will focus on two directions: (i) Developing a new method to the study of selection on complex human traits-built upon a marriage between the statistical power of standard GWAS and the immunity of family studies to various GWAS confounders; and (ii) Understanding how variants with genetic effects contingent on the environment evolve.

The history of environment-specific selection is hypothesized to have been highly consequential for human health. We will develop new theory and a matching statistical inference tool to understand selective constraint on gene-by- environment interactions, and their consequences for contemporary genetic architecture of complex diseases.