Novel statistical genetics methods to unravel polygenic interactions in complex traits

Enter the text here that is the new abstract information for your application. Complex traits result from interactions between many genetic and environmental factors. Nonetheless, most complex trait studies assume an additive model, in which genetic effects are independent of the environment and each other. This simple model has successfully identified many trait-associated loci, and these loci can be combined into Polygenic Scores (PGS) to predict disease. However, these results have not generally identified novel disease biology or therapies. Worse yet, current PGS have known and unknown statistical biases that will impede their accuracy in the clinic. I hypothesize that genetic interactions are the missing link in our understanding of complex trait biology. Genetic interactions are central to many fields of biology, and it is not likely that complex human traits are fundamentally different. However, prior studies of genetic interactions have generally been unsuccessful. I argue this results from limitations in our current models. In the next five years, I will develop genetic interaction models for complex traits to address these limitations. First, I will develop models to identify gene-gene interaction at the level of pathway-pathway interaction that build on my recent “Coordinated” framework for epistasis. Coordination is biologically plausible and statistically powerful. I will extend my Coordinated models to decompose pleiotropic effects on multiple traits and to unravel subtypes of common diseases. Second, I will develop rigorous and powerful models of gene-environment interaction that apply to novel areas of complex trait genetics. I will study cell type-specific heritability in single cell ‘omics data, I will incorporate context-specific effects to improve power and portability in PGS, and I will quantify the heritability of treatment response from biobank data. My methods will be mathematically rigorous and computationally efficient. They will build on my track record of developing robust and freely-distributed statistical genetics methods. I will apply my methods to phenome-wide scans in several large-scale cohorts, especially to ensure the PGS predictions are rigorous and replicable. I will also study Major Depressive Disorder in detail, a classic example of a heterogeneous complex disorder with a mix of poorly understood genetic and environmental causes. My interaction methods will close the gap between statistical explanation and biological understanding, revealing new paths to precision medicine that benefit everyone.