To illustrate the method, we applied gene prediction models (derived from whole blood) consisting of DGN elastic net predictors to the seven WTCCC disease studies -- bipolar disorder (BD), coronary artery disease (CAD), hypertension (HT), type 1 diabetes (T1D), type 2 diabetes (T2D), Crohn’s disease (CD), and rheumatoid arthritis (RA)22 . Genotypes imputed to the 1000G reference sets were used. Imputation was done using the University of Michigan Imputation-Server and the same parameters as described for the imputation of DGN data. For each disease, cases and controls (1958 Birth Cohort and the UK Blood Service Cohort) were jointly imputed to avoid subtle differences between cases and controls not attributable to disease risk. We excluded all SNPs with an imputation R2 < 0.8 and for computational speed we kept only the HapMap Phase II subset of SNPs.
