Prediction R-squared evaluated in testing data sets (average over 30 randomly drawn testing data sets, each having 500 individuals) by training and validation data sets and model.

<p>N-FHS = Number of records from Framingham, N-GEN = Number of records from GENEVA. G-BLUP uses 400 K SNPs, wG-BLUP uses 400 K SNPs, but the contribution of each SNP to the genomic relationship matrix was weighted using as weight, where is the SNP associated p-value reported by <a href="http://www.plosgenetics.org/article/info:doi/10.1371/journal.pgen.1003608#pgen.1003608-LangoAllen1" target="_blank">[5]</a>.