Association graph reduction: a framework for efficient exploration of additive and interactive effects in complex disease etiology

Jeffrey Kilpatrick, Rice University

Photo of Jeffrey Kilpatrick

Many interesting biological traits are caused by a combination of environmental exposure and genetic predisposition. While sufficient characterization of transient exposure for the purposes of elucidation of phenotype etiology remains difficult or impossible for many organisms, studying the unchanging genetic component of risk seems feasible. To this end, many researchers have attempted to identify so-called susceptibility loci (genetic polymorphisms which modify risk) by testing sets of markers spread throughout subjects’ genomes. Unfortunately, this strategy has proved to be more difficult than initially expected. The greatest problem with this or any approach is the fact that many polymorphisms may collude either independently or interactively to modify risk. Worse, these factors may vary among subjects. We propose association graph reduction (AGR), a framework which, unlike existing methods, acknowledges this complexity to successfully identify susceptibility loci. We compare AGR to several existing methods and show it is extremely powerful and computationally efficient.

Abstract Author(s): Jeffrey R. Kilpatrick and Luay K. Nakhleh