Initially, we sought to provide a preview of the roadmap ahead using analytical derivation. In light of upcoming large-scale genetic studies, this approach will also set expectations for opportunities and limitations for future genetic risk prediction studies of PTSD. These projections are predicated on standard assumptions and models from quantitative genetic theory (Supplementary Materials). Using the heritability estimate of 30% (obtained from early male twin studies and recent SNP heritability estimates for women) and an estimated disease prevalence of 8%, the optimal panel trained on an infinite number of samples would have an AUC of a little over 80% (Fig. 1). It should be noted that unlike most study samples, including the present study samples, where cases are intentionally oversampled so as to make up half of a study cohort (i.e., ascertainment), both the training sample (whose sample size is shown in the horizontal axis) and replication sample (whose performance is shown in the vertical axes) are assumed to be drawn randomly and independently from the general public where disease incidence rate is 8%.
