Fisher’s exact tests were used to examine the overlap in the top associated SNPs across all pairwise combinations of neuroticism items and the sum-score. Prior to this analysis, we applied LD-based pruning on the summary statistics from all items to ensure that SNPs were independent (a subset of 10,000 UKB participants was used as a reference set; r2 = 0.8). Sign tests were used to establish whether signs of top associated SNPs were in the same direction across all neuroticism items and the sum-score. Here we used clumping (r2 = 0.1 with window size 500 kb) based on the association P values to ensure that the most strongly associated variants were not lost. As clumping results in different SNP sets for different items, sign test results are not symmetrical for pairs of items. For both tests, we used multiple P value thresholds, including the conventional GWS threshold (5 × 10−8) and a conservative threshold correcting for the 13 phenotypes tested (3.85 × 10−9).
