Mapping variants to amino-acid changes in three-dimensional protein space improves aggregate association test power and suggests mechanisms of action
Year: 2016
Type: other
Abstract: To explore rarer variants than those identified by GWAS studies of complex traits such as Type 2 Diabetes (T2D), aggregate tests have been proposed to increase power. As aggregate analysis is negatively impacted by the inclusion of non-functional variants, the majority of aggregate associations identified to date have tested variants that truncate the protein (PTVs), or are functional in cellular assays. However, PTVs exclude the majority of variants in a gene, cellular assays are seldom available, and bioinformatics algorithms have been shown to be poor predictors of cellular function. We hypothesized that three-dimensional (3-D) protein structure maps might enable more accurate groupings of variants for aggregate association tests. We developed a novel method to (a) map all missense mutations to 3-D structures of proteins; (b) identify patterns of amino-acid changes in 3-D space; and (c) cluster variants according to these patterns for use in aggregate association analysis. We applied our methodology to analyze missense variants in TBC1D1 and TBC1D4 as identified from sequence data of 16,857 multi-ethnic T2D cases/controls; we chose these genes because TBC1D4 was recently identified to carry a Greenlandic variant of large effect on T2D risk and has an established role in insulin signaling, while TBC1D1 has a potentially (but unverified in humans) analogous role to TBC1D4 based on animal models. PTVs in TBC1D1 (OR=2.14, p=0.02), as well as combined PTVs and bioinformatically deleterious variants in TBC1D4 (OR=1.2, p=0.03), demonstrated nominally significant aggregate association with increased T2D risk in the heterozygous state. However, bioinformatically deleterious missense variants in TBC1D1 showed no association (OR=1.02, p=0.88). In contrast, our mapping algorithm identified a cluster of mutations localized on the surface of the TBC1D1 Rab-GAP binding site, which in aggregate demonstrated a nominal association with T2D risk (OR=2.7, p=0.038). Rab-GAP plays a known role in localization of the glucose transporter GLUT4 to the plasma membrane in insulin responsive tissues, with TBC1D1 and TBC1D4 binding necessary for activation of Rab. Collectively, our results thus suggest (a) that disruption of TBC1D1 in humans increases risk for T2D; (b) that impaired Rab-GAP binding is responsible for this increase in risk; and (c) that analyzing variants in 3-D protein space can inform aggregate tests of missense variants when cellular assay data is unavailable.
Institution: Broad Institute
Language: English
Subfield: Genetics
Domain: Life Sciences
Open Access status: gold
OpenAlex ID: W2583934158