A key challenge for genetic analysis today is to account for the bulk of the phenotypic variance in complex traits attributable to genetic factors. Traditional genetic analysis methods exploit the possibility that variation in a single genetic locus may result in detectable effects on a phenotype. These methods typically examine genotypes one at a time, and build additive (or log-additive) effects models from them. The approach has been successful, with genome-wide association studies (GWAS) alone identifying over 500 genetic variants contributing to disease (Hindorff et al. 2009). However, for many complex traits (e.g. obesity, smoking, diabetes), the variants identified by studies with large samples and dense genome-wide geno-typing account for only a modest fraction of the phenotypic variance estimated to be attributable to genetic contributions (Goldstein 2009; Hirschhorn 2009; Kraft and Hunter 2009). A variety of factors have been suggested to account for the “missing variance”, including rare variants, variants not surveyed by current GWAS chips, structural variants (e.g. copy number variants such as insertion/deletions or copy neutral variation such as inversions and translocations), population heterogeneity, gene–gene interactions, and gene-environment interactions (Galvan et al. 2010; Manolio et al. 2009).
