We developed a GPU-based analytical method, named as SHEsisEpi, which purely focuses on risk epistasis in a genome-wide association study (GWAS) of complex traits, excluding the contamination of marginal effects caused by single-locus association. We analyzed the Wellcome Trust Case Control Consortium's (WTCCC) GWAS data of bipolar disorder (BPD) with 500K SNPs. Our algorithm only used 27 h to finish the exhaustive scan and was more than 300 times faster than the CPU-based analysis on our system. Furthermore, by genotyping the top finding that met our criteria (P = 5.37 × 10−12) and its nearby SNP pairs in another independent 475 BPD patients and 480 normal controls from Chinese Han population, we validated these findings, related with two gene pairs, conferring risk (α = 0.05) to BPD. Binary files and source codes of our program can be downloaded at http://analysis.bio-x.cn from the main menu of SHEsis.

GWAS, which is a simplistic, exhaustive, whole genome 'one SNP at a time' analysis approach focused on single-locus marginal effects 1, has led to the discovery of a lot of risk loci of complex diseases. But in most cases, there are more than one pathogenic loci. Much empirical evidence has shown that complex traits are broadly affected by interactions among different loci 2, 3, 4, 5. Theoretically there are a variety of purely epistatic disease models, which display no main effects at all 6. In this occasion, genome-wide gene-gene interaction analysis is requested to identify hidden multi-locus susceptibility, especially risk interaction effects without contamination of single-locus association.

Log in or create a free account to read this content

Gain free access to this article, as well as selected content from this journal and more on nature.com

Access through your institution

or

Sign in or create an account
Continue with Google
Continue with ORCiD