In the ISC report, the P-value of the polygenic score decreases as the SNP inclusion threshold increases. This seems to suggest that a large number of associated markers lie within the mass of individually non-significant SNPs. In Figure 1 and Figure 2, the expected P-value of the polygenic score is shown as a function of the inclusion threshold, with the explained genetic variance set to its estimated value of 28.7% and other parameters as stated above. The figures show that this trend could be observed when as many as 90% of SNPs have no effects, and for the linear regression estimator the significance of the score continues to improve until the whole marker panel is included. Only for a very high proportion of null SNPs is there an optimal inclusion threshold less than 1. The allele count estimator has an optimum threshold less than 1 for all scenarios, but it is consistently less significant. Thus, in this dataset with high power, decreasing P-values are consistent with a range of polygenic models including those with a high proportion of null markers.
