PheWAS for human data was performed using 29,722 individuals with Illumina HumanExome array data identified as European ancestry in the EHR and by using structure54. To define diseases, we mapped International Classifications of Diseases, 9th edition, (ICD9) codes from the EMR into 1,645 possible PheWAS phenotypes using methods described previously6. PheWAS phenotypes aggregate like ICD9 codes together (for example, type 2 diabetes codes as a specific phenotype), are hierarchical (for example, ‘inflammatory bowel disease' is a parent of ‘Crohn's disease' and ‘Ulcerative colitis'), and include logic to select controls for each case definition. We considered only phenotypes with at least 20 individuals for analysis, and required each case to have at least two ICD9 codes for a PheWAS phenotype to be considered a case (those with only one code are neither a case nor a control). Each SNP-phenotype association test was run with PLINK (ref. 55) using logistic regression adjusted for age, sex and the first three principal components as calculated by EIGENSTRAT using ancestry informative markers. Analysis was performed assuming an additive genetic model. These data were aggregated and
