We also compared the 10-fold cross-validated prediction R2 from elastic net models with different starting SNP sets from the DGN genotype imputation (4.6M 1000 Genomes Project SNPs (MAF>0.05, R2>0.8, non-ambiguous strand), the 1.9M of these SNPs that are also in HapMap Phase II, and the 300K of these SNPs that were genotyped in the WTCCC).
