For each WTCCC disease, we estimated GReX^EN, and tested it for association with disease risk using logistic regression in R (R-project.org). We restricted our PrediXcan analysis to include genes with a cross-validated prediction R2 > 0.01 (10% correlation) in the DGN sample. Because the WTCCC studies use shared controls, pleiotropy analyses using these datasets would not be straightforward, and comparison of results across diseases was avoided.
