Fig. 2.
Ability of polygenic scores for cerebral palsy (CP) and seven related traits to distinguish between individuals with and without CP in two independent cohorts. A) Single-PGS logistic regressions, adjusted for sex and 10 ancestry PCs. Star = significant association of polygenic score with CP status after p-value adjustment for eight comparisons. B) Multiple-PGS logistic regression (all eight polygenic scores in the same model), adjusted for sex and 10 ancestry PCs. Star = significant association of polygenic score with CP status after p-value adjustment for eight comparisons. C) Area under the receiver operating characteristics curve for each polygenic score, or all eight polygenic scores in the same model. Star = AUC significantly >0·5, after p-value adjustment for nine comparisons. D) Variance explained in CP on the liability scale attributable to each polygenic score, or to all eight polygenic scores in the same model. Population prevalence of 0·0014 was used for the Australian cohort (1 in 700 live with CP in Australia) and 0·0032 used for the MyCode cohort (1 in 313 live with CP in the U.S.). Variance on the liability scale explained significantly >0 after p-value adjustment for nine comparisons is indicated by ∗. Points = means; bars = 95% CIs for A–B (two-tailed Wald tests, α = 0·05), and 90% CIs for C–D (one-tailed Z-tests, α = 0·05). Dark blue circles = full Australian cohort (ncases = 525, ncontrols = 20,410), teal triangles = MyCode cohort with cases and controls matched on age and sex (ncases = 322, ncontrols = 1610).