Fig. 1.

Workflow for polygenic score analysis. Discovery GWAS summary statistics for CP and seven related traits were used to generate polygenic weights. These traits included a trait related to motor development, and known CP predisposing factors and comorbidities with sufficiently powered GWAS (≥5 independent genome-wide significant variants). Polygenic weights were applied to two independent target cohorts to calculate individual polygenic scores for each trait. Polygenic scores predictive performance was evaluated using logistic regression, area under the receiver operating characteristic curve (AUC), and variance explained in CP on the liability scale (R²_l). The primary analysis used the full Australian cohort and the age-matched MyCode cohort. Another primary analysis was only conducted in the Australian cohort and stratified the cohort by the presence or absence of a monogenic diagnosis for CP. A sensitivity analysis included the youngest 525 Australian controls (1:1 with cases) to assess age bias, and the full, unmatched MyCode cohort.