Simulating systemic effects of expression quantitative trait loci across gene regulatory networks
Authors: Matthew Aguirre 1; Guy Sella 2,3; Jonathan Pritchard 4,5
Affiliations: 1) Department of Biomedical Data Science, Stanford University, Stanford, CA; 2) Department of Biological Sciences, Columbia University, New York, NY; 3) Program for Mathematical Genomics, Columbia University, New York, NY; 4) Department of Genetics, Stanford University, Stanford, CA; 5) Department of Biology, Stanford University, Stanford, CA
Keywords: Complex traits
Gene regulatory networks (GRNs) govern many of the core developmental and biological processes which give rise to complex traits. Even as genome-wide transcriptomic resources approach population scale, it remains challenging to interpret how the structure of GRNs impacts the distribution of genetic effects on gene expression. Learning the genetic architecture of gene expression traits is a key aim in quantitative genomics, but there is still an unmet need for theoretical modeling which places the statistical basis of these traits in the mechanistic context of GRNs. Here, we propose a simple approach to model and simulate the structure and function of GRNs, making use of techniques from small world network theory and dynamical systems models of gene regulation. Specifically, we model gene expression regulation using a stochastic differential equation with terms for endogenous transcription and contributions from parent nodes (transcription factors) in a GRN. This formulation permits variation in regulatory parameters which naturally corresponds to effects from expression quantitative trait loci (eQTLs) or intracellular interventions (e.g., targeted gene knock-down or knock-outs). We use this model to generate synthetic samples of gene expression data, and probe the system-wide effects of perturbing key regulators in the GRN. We further this analysis by simulating a population of individuals which harbor a substantial burden of eQTLs, investigating the distribution of gene expression heritability introduced by these factors in the network. We conclude by discussing implications of our work towards understanding the architecture of heritable genetic variation in complex traits, even beyond gene expression.