Whole-genome sequencing to understand the genetic architecture of common gene expression and biomarker phenotypes.
Human Molecular Genetics 2014 ; 24: 1504-12.
Wood AR, Tuke MA, Nalls M, Hernandez D, Gibbs JR, Lin H, Xu CS, Li Q, Shen J, Jun G, Almeida M, Tanaka T, Perry JR, Gaulton K, Rivas M, Pearson R, Curran JE, Johnson MP, Göring HH, Duggirala R, Blangero J, McCarthy MI, Bandinelli S, Murray A, Weedon MN, Singleton A, Melzer D, Ferrucci L, and Frayling TM
DOI : 10.1093/hmg/ddu560
PubMed ID : 25378555
PMCID : PMC4321449
URL : https://academic.oup.com/hmg/article-lookup/doi/10.1093/hmg/ddu560
Initial results from sequencing studies suggest that there are relatively few low-frequency (<5%) variants associated with large effects on common phenotypes. We performed low-pass whole-genome sequencing in 680 individuals from the InCHIANTI study to test two primary hypotheses: (i) that sequencing would detect single low-frequency-large effect variants that explained similar amounts of phenotypic variance as single common variants, and (ii) that some common variant associations could be explained by low-frequency variants. We tested two sets of disease-related common phenotypes for which we had statistical power to detect large numbers of common variant-common phenotype associations-11 132 cis-gene expression traits in 450 individuals and 93 circulating biomarkers in all 680 individuals. From a total of 11 657 229 high-quality variants of which 6 129 221 and 5 528 008 were common and low frequency (<5%), respectively, low frequency-large effect associations comprised 7% of detectable cis-gene expression traits [89 of 1314 cis-eQTLs at P < 1 × 10(-06) (false discovery rate ∼5%)] and one of eight biomarker associations at P < 8 × 10(-10). Very few (30 of 1232; 2%) common variant associations were fully explained by low-frequency variants. Our data show that whole-genome sequencing can identify low-frequency variants undetected by genotyping based approaches when sample sizes are sufficiently large to detect substantial numbers of common variant associations, and that common variant associations are rarely explained by single low-frequency variants of large effect.