%0 Journal Article %A Gang Wang %A Ence Yang %A Jizhou Yang %A Beiyan Zhou %A Yanan Tian %A James J. Cai %T Epistasis and decanalization shape gene expression variability in humans via distinct modes of action %D 2015 %R 10.1101/026393 %J bioRxiv %P 026393 %X Increasing evidence shows that phenotypic variance is genetically controlled; the precise mechanisms of genetic control over the variance remain to be determined. Here, using variance association mapping analysis of gene expression, we show that common genetic variation contributes to gene expression variability via distinct modes of action—e.g., epistasis and decanalization. We focused on the full set of genetic loci associated with gene express variance, i.e., genome-wide expression variability QTLs (or evQTLs), in humans. We found that a quarter of evQTLs, explained by the epistasis model, could be attributed to the presence of “third-party” partial eQTLs that influence gene expression in a fraction, rather than the entire set, of samples. The other three-quarters of evQTLs, explained by the decanalization model, were attributable to the disruptive effect conferred by their own SNPs on transcriptional robustness—that is, these SNPs are responsible for modulating the stability of transcriptional machinery. To validate the model, we measured the discordant expression between monozygotic twins, as well as the level of transcriptional noise in individual cell lines. We showed that decanalizing evQTL SNPs indeed affect the level of transcriptional noise in individuals and contribute to gene expression variability at the population level. Together, our results suggest that common genetic variation works either interactively or independently to influence gene expression variability. These findings may have implications for methodology development toward a new variability-centered research paradigm for mapping quantitative traits. %U https://www.biorxiv.org/content/biorxiv/early/2015/12/14/026393.full.pdf