Testing the performance of polygenic scores for multiple traits to explain cerebral palsy in two independent cohorts
Files
(Published version)
Date
2026
Authors
Thomas, J.T.
Berry, A.S.F.
Oetjens, M.T.
Berry, J.G.
MacLennan, A.H.
Gordon, S.D.
Hale, A.T.
Olsen, C.M.
Whiteman, D.C.
Torene, R.I.
Editors
Advisors
Journal Title
Journal ISSN
Volume Title
Type:
Journal article
Citation
EBioMedicine, 2026; 126:106208-1-106208-13
Statement of Responsibility
Jodi T. Thomas, Alexander S. F. Berry, Matthew T. Oetjens, Jesia G. Berry, Alastair H. MacLennan, Scott D. Gordon, Andrew T. Hale, Catherine M. Olsen, David C. Whiteman, Rebecca I. Torene, David H. Ledbetter, Nicholas G. Martin, Clare L. van Eyk, Jozef Gecz, Scott M. Myers, Brittany L. Mitchell, and Mark A. Corbett
Conference Name
Abstract
Background: Cerebral palsy (CP) is a complex neurodevelopmental disorder with both environmental and genetic contributors. Rare genetic variants explain a substantial proportion of CP, but the contribution of common variants remains unclear. We evaluated whether polygenic scores for CP and related traits explain the aetiology of CP. Methods: We analysed two independent target cohorts: a case–control cohort including people with a confirmed clinical diagnosis of CP from the Australian CP Biobank and population-based controls; and MyCode, a United States healthcare cohort with CP status identified by electronic health records. Only participants of European genetic ancestry were included. CP polygenic scores were constructed using a publicly available discovery genome-wide association meta-analysis of Finnish and UK cohorts (ncases = 624, n controls = 495,687) and applied to the target cohorts for out-of-sample prediction. Additional polygenic scores were generated from publicly available genome-wide association studies for seven CP-related traits. Predictive performance was assessed using logistic regression, area under the receiver operating characteristic curve, and variance in CP liability explained. Findings: The Australian cohort included 525 cases and 20,410 controls, and MyCode 322 cases and 1610 agematched controls. The combined model of all eight polygenic scores significantly discriminated CP status, explaining 1⋅3% of CP liability in the Australian cohort (90% CI lower bound 0⋅82%, padj<0⋅0001), and 0⋅78% in MyCode (90% CI lower bound 0⋅35%, padj<0⋅0001). CP-specific polygenic scores demonstrated minimal predictive signal, likely reflecting limited GWAS power. Polygenic scores for known CP predisposing factors (birth weight, gestational duration, stroke) showed modest predictive performance, with some cohort differences. Results were similar when the Australian cohort was stratified by monogenic CP diagnosis. Interpretation: Our findings demonstrate a measurable polygenic contribution to CP and shared genetic influences with predisposing factors, including those traditionally considered environmental, and comorbidities. Common variants appear to contribute broadly to CP susceptibility, highlighting a multifactorial landscape relevant for earlier diagnosis and intervention.
School/Discipline
Dissertation Note
Provenance
Description
Access Status
Rights
© 2026 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).