We evaluated the performance of TWAS by identifying significantly associated genes in the 2010 lipid study that did not overlap a genome-wide significant SNP, and looking for newly genome-wide significant SNPs in the expanded 2013 study. The P-value for the number of genes with increased significance and genome-wide significance in the 2013 study was computed by a hypergeometric test, with background probabilities estimated from the set of significantly heritable genes. Of the genes not overlapping a significant locus in the 2010 study, 70% had a more significant SNP in the 2013 study and 3.5% overlapped a genome-wide significant SNP (P<5×10−08).
