In this work we proposed methods that integrate genetic and transcriptional variation to identify genes with expression associated to complex traits. Using imputed gene expression to guide GWAS has three potential advantages. First, the gene is a more interpretable biological unit than an associated locus, which often contains multiple significant SNPs in LD that may not lie in genes and/or tag variants in multiple genes. Second, the lower total number of genes (or cis-heritable genes) means the multiple-testing burden is substantially reduced relative to all SNPs. Lastly, combining cis-SNPs into a single predictor may capture heterogeneous signal better than individual SNPs or cis-eQTLs. Focusing the prediction on the genetic component of expression also avoids confounding from environmental differences caused by the trait that may impact expression. Our approach builds upon the wealth of GWAS data in massive cohorts to directly implicate the gene-based mechanisms underlying complex traits.
