Second, the accuracy of the PRS is dependent on the informativity of the discovery sample which is used to define the disease-associated variants and risk alleles. Studies with larger sample sizes yield more power to detect small genetic effects and provide more accurate estimates of the effect sizes of SNPs 22. Adequate sample sizes today typically include tens of thousands of cases and a similar or larger number of controls. Sample sizes, and consequently power, are expected to increase over time going forward. However, large sample sizes do inherently carry increased heterogeneity at both clinical and genetic levels and have the potential of making genetic association signals less clear. Samples for large collaborative efforts are often ascertained across multiple investigator sites and in multiple countries, where diagnostic practices and exposure to demographic/epidemiological risk factors might be slightly different, potentially introducing artifactual associations and contributing to fluctuating strengths of association and effect size estimates across cohorts. In general, PRS based on samples which are from the Psychiatric Genomics Consortium and for which summary statistics are publically available have been reliably vetted and processed (but currently are based largely on individuals of Caucasian European descent, the implications of which are described above).
