pgen.1009060.s007.tif (1.1 MB)

Comparison of CT-TWAS aggregation methods.

Download (1.1 MB)
posted on 15.12.2020, 18:27 by Corbin Quick, Xiaoquan Wen, Gonçalo Abecasis, Michael Boehnke, Hyun Min Kang

Comparison of Cross-Tissue TWAS (CT-TWAS) p-values, and p-values using only the top single tissue, for disorders of lipoid metabolism using GWAS summary statistics from the UK Biobank. The top tissue was defined as the tissue with the largest number of significant genes using FWER threshold α = 0.05 with Bonferroni adjustment for the number of eGenes in each tissue. In this case, the top tissue was “Liver” with 27 significant genes out of 3,314 total eGenes (Bonferroni-adjusted p-value threshold = 1.5 × 10−5). Top-Tissue p-values are compared with CT-TWAS p-values (CT-Q, CT-A, and CT-M), which aggregate across all 47 tissues, restricted to Liver eGenes. CT-Q is calculated using the sum of squared single-tissue TWAS z-scores (similar to SKAT); CT-A is calculated by combining single-tissue TWAS p-values using ACAT; and CT-M is calculated from the minimum single-tissue p-value using the multivariate normal joint density of all single-tissue z-scores (described in Materials and methods). Here, CT-M detected 51 significant genes, followed by CT-A with 47, CT-Q with 33, and top-tissue-only with 27.