prediction32,33, a substantial portion of variance in expression can be predicted at current sample sizes due to the much smaller number of independent SNPs in the cis region21. Furthermore, larger training sizes will continue to increase the total number of genes that can be accurately predicted (Figure 3). We further evaluated cross-cohort prediction of these genes in the YFS and NTR, which were roughly equally sized and had expression measured in whole-blood by microarray, but were genotyped on different platforms and from different Scandinavian populations. After accounting for cis-heritability in the test cohort, our cross-cohort standardized accuracy (i.e. r2/hg,cis2) was broadly consistent with in-cohort cross-validation accuracy (Supplementary Table 3). The BSLMM was again the most accurate predictor, with an average cross-cohort r2/hg,cis2 of 72%, outperforming the best eQTL by an average 1.17x.
