Abstract
Quantifying the overall magnitude of every single locus’ genetic effect on the widely measured human phenome is of great challenge. We introduce a unified modelling technique that can consistently provide a total genetic contribution assessment (TGCA) of a gene or genetic variant without thresholding genetic association signals. Genome-wide TGCA in five UK Biobank phenotype domains highlights loci such as the HLA locus for medical conditions, the bone mineral density locus WNT16 for physical measures, and the skin tanning locus MC1R and smoking behaviour locus CHRNA3 for lifestyle. Tissue-specificity investigation reveals several tissues associated with total genetic contributions, including the brain tissues for mental health. Such associations are driven by tissue-specific gene expressions, which share genetic basis with the total genetic contributions. TGCA can provide a genome-wide atlas for the overall genetic contributions in each particular domain of human complex traits.
Highlights
Π−k, and the + and − subscripts indicate that the Z-scores were generated from positive and negative components of the mixture, respectively
Similar to the case with independent Z-scores, for each true model, denoting a set of generated correlated Z-scores as fzÀ1; Á Á Á ; zÀkÀ ; z01; Á Á Á ; z0k0 ; zþ[1]; Á Á Á ; zþkþ g, where k0 = π0k, k+ = π+k, k− = π−k, and the + and − subscripts indicate that the Z-scores were generated from positive and negative components of the mixture, respectively
We split the SNPs into 100 subsets, where each subset contained SNPs j, j + 100, j + 200, ⋯, j = 1, 2, ⋯, 100, so that the LD correlations were pruned. These resulted in 100 ^γ estimates, and we report the median of them
Summary
We first applied the mixture model to the GWAS summary statistics of 1376 UKBB phenotypes and estimated and tested genome-wide TGCA Θ of 2,029,920 quality-controlled SNPs with minor allele frequencies (MAFs) > 0.005. We examined the squared Z-score (χ2) distribution for Θ^ in each phenotype domain, stratified on the cis-eQTL for expressed genes in each tissue (Fig. 4, Supplementary Fig. 14). As described above, such an investigation might not be trustworthy for lifestyle and diet phenotypes, while evidence could be found for the other phenotype domains. Such signals were enriched for physical measures in colon sigmoid
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
