Association and replication of SNP-SNP interactions for hundreds of gene expression phenotypes. A. Fish, W. Bush Center for Human Genetics Research, Vanderbilt University, Nashville, TN.

   SNP-SNP interactions are thought to play a major role in the etiology of many complex phenotypes; however, they have not been extensively investigated due to computational and statistical limitations, such as the large number of tests necessary and the difficulty in replicating complex genetic effects. Gene expression is an extremely intricate process, with multiple protein complexes interacting with DNA to facilitate transcription. The role of gene regulation in the etiology of disease been increasingly recognized, as the majority of significant findings from GWAS occur in gene regulatory regions. Correspondingly, SNP-SNP interactions within the cis-regulatory regions may play an important role in disease susceptibility. In this study, we investigated interactions been single nucleotide polymorphisms (SNPs) affecting gene expression in 210 HapMap samples with both genotype information and gene expression data in lymphoblastoid cell lines. We limited our analysis to pairs of SNPs which were previously identified as expression quantitative trait loci (eQTLs) with nominal significance (p < 0.05). Amongst these eQTLs with significant main effects, we created all possible pair-wise combinations for each gene, resulting in a total of approximately 30 million tests. We created models containing both additive and dominant components to avoid a systematic bias inherent to standard multiplicative interaction terms, in which deviations from additivity in single SNP effects are falsely identified as significant interaction terms. Using a false discovery rate (FDR) correction level of 0.05, we identified 4,284 significant interactions, which corresponded to 123 distinct genes. Using an additional set of 466 multiethnic HapMap samples, we investigated 596 of the initial 4,284 significant interactions. Of these 596 interactions, 281 replicated using a FDR correction of 0.05. Functional annotation of the specific SNPs was also performed, in order to determine potential biological mechanisms which might be predisposed to gene-gene interactions. These results indicate that non-linear interactions among cis-regulatory variants may be a common component to the architecture of gene expression.

You may contact the first author (during and after the meeting) at