Analysis of variants obtained through whole-genome sequencing provides an alternative explanation to apparent epistasis. A. R. Wood1, M. A. Tuke1, M. Nalls2, D. Hernandez2,3, S. Bandinelli4,5, A. Singleton2, D. Melzer6, L. Ferrucci7, T. M. Frayling1, M. N. Weedon1 1) Genetics of Complex Traits, University of Exeter Medical School, Exeter, United Kingdom; 2) Laboratory of Neurogenetics, National Institute of Aging, Bethesda, Maryland, USA; 3) Department of Molecular Neuroscience and Reta Lila Laboratories, Institute of Neurology, UCL, London, United Kingdom; 4) Tuscany Regional Health Agency, Florence, Italy, I.O.T. and Department of Medical and Surgical Critical Care, University of Florence, Florence, Italy; 5) Geriatric Unit, Azienda Sanitaria di Firenze, Florence, Italy; 6) Institute of Biomedical and Clinical Sciences, University of Exeter Medical School, Barrack Road, Exeter, United Kingdom; 7) Longitudinal Studies Section, Clinical Research Branch, Gerontology Research Center, National Institute on Aging, Baltimore, Maryland, USA.

   It has proven hard to find examples of gene-gene interaction (epistasis) in humans. The first evidence of epistasis affecting traits was recently described by Hemani et al. (Nature, 2014). They detected 30 pairwise interactions influencing gene expression levels that were replicated in additional studies. We sought further replication but used genotypes derived from low-pass whole-genome sequencing to capture more completely the variation around the putatively interacting variants. Using 450 unrelated individuals from the InCHIANTI study with genome-wide expression profiling captured on the Illumina HT-12 expression microarray from whole-blood, we replicated 14 of the reported pairwise interaction effects (P<0.05). However, in each case, a third variant captured by whole-genome sequencing could explain all of the apparent epistasis in our data. This third variant was often moderately correlated with each of the two putatively interacting variants, despite very low levels of correlation between the original pair. For example, evidence for putative interactions between pairs of SNPs in cis influencing FN3KRP (P=3x10-12), CSTB (P=8x10-07), MBLN1 (P=3x10-06) and NAPRT1 (P=6x10-06) disappeared on correction for a single confounding sequenced variant (FN3KRP (P=0.43), CSTB (P=0.99), MBLN1 (P=0.16) and NAPRT1 (P=0.84)) Our results provide an alternative explanation for the apparent epistasis observed for gene expression traits in humans.

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