use in PRS, so that PRS can be computed as a sum of weighted genotypes). An example of a population distribution of a PRS, showing a normal distribution, is illustrated in Figure 1. PRS can be calculated using different sets of disease-associated variants, and typically different P value thresholds for disease association are used to create a series of PRSs for a particular disease or trait. The P value threshold at which the best distinction is observed between case-control groups (or with variability in a quantitative trait) is selected as optimal. Intuition dictates that only variants that are robustly associated at genome-wide significance contain genuine disease risk predictors and no “noise”; however, such predictors generally explain very little variation in disease risk and therefore have little predictive accuracy 6. The common practice, therefore, is to perform genetic prediction by using many independent risk alleles across the genome—including many hundreds of thousands of SNP variants, most of which show only weak evidence of disease association on their own—under the assumption that many genuine significant associations are potentially missed because of inadequate power in the original GWAS. This approach yields greater predictive power across many psychiatric disorders, maximizing the predictive power and
