analysis7,21 (Figure 4). The OR for being a case per standard deviation (SD) increase in PGS was 1.57. The OR for being a case in the tenth compared with the first decile of PGSs was 4.92 (95% CI 4.57–5.29) (Figure 4), and the OR for the top versus bottom centiles was 11.8 (95% CI 8.4–15.2) (Figure 4). Heterogeneity in the out-of-sample prediction results could be partly explained by the recorded ascertainment type (Figure 4; key resources table), which we classified as “clinical” (12 cohorts; ascertained from in- or out-patient settings or EHR) or “community” ascertained (30 cohorts; interviews or questionnaires self-reporting on lifetime depression). The difference in mean PGS between clinical versus community cases was 0.131 (SE 0.012, p < 2 × 10−16) control sample SD units. The non-linear shape of these decile plots is expected under a polygenic architecture.22
