The International Multiple Sclerosis Consortium performed a similar exercise using a training sample of 931 cases and 2431 controls, a replication sample of 876 cases and 2077 controls, and a marker panel of 59470 nearly independent SNPs [6]. They also observed decreasing P-values for association as more SNPs were included in the score, obtaining when all SNPs were included. Assuming prevalence of 0.1% this analysis has 80% power at nominal significance for explained genetic variance of 9.4%, and the observed result yields an estimate of 31.5% (95%CI: 24.9%–37.9%) assuming all SNPs have effects.
