We employed TWAS to identify expression-trait associations at the 697 known GWAS risk loci for height7 using the YFS data for which height was also measured. At each locus, we considered three strategies for selecting a single causal gene: 1) the gene nearest to the top GWAS SNP; 2) the gene for which the index SNP is the strongest eQTL in the training data; 3) the most significant TWAS gene. For each strategy, we then constructed a risk-score using the genetic value of expression for the selected genes and correlated the risk score with height measurements in the YFS individuals (an independent sample from the original height GWAS, Supplementary Note). The R2 between the risk score and height was 0.038 (nearest); 0.031 (eQTL); and 0.054 (TWAS); with TWAS significantly higher than the others in a joint model (Supplementary Table 5, Methods). When we re-computed the risk scores using TWAS values for expression from the NTR cohort (which introduces additional noise due to heterogeneity between the cohorts), TWAS remained significantly higher than the eQTL strategy, but was comparable to selecting the
