Method

Transfer learning enhances clinical utility of polygenic scores with small, phenotypically refined cohorts

    • 1Division of Biostatistics, Dalla Lana School of Public Health, The University of Toronto, Toronto, M5T 3M7, Canada;
    • 2Department of Neurology, Odense University Hospital, Odense 5000, Denmark;
    • 3Department of Neurology, Second Faculty of Medicine, Charles University, Motol and Homolka University Hospital, Prague 150 06, Czech Republic;
    • 4Department of Neurology, Cardiff and Vale University Health Board, Cardiff CF14 4XW, United Kingdom;
    • 5Department of Women's and Children's Health, Karolinska Institute, Stockholm 171 77, Sweden;
    • 6Department of Pediatric Neurology, Karolinska University Hospital, Stockholm 171 76, Sweden;
    • 7Institute of Biological Sciences, Faculty of Science, University of Malaya, Kuala Lumpur 50603, Malaysia;
    • 8Adult Genetic Epilepsy Program, Krembil Research Institute, University of Toronto, Toronto M5T 0S8, Canada;
    • 9Department of Neurology, Drammen Hospital, Vestre Viken Health Trust, Oslo 3004, Norway;
    • 10Department of Epilepsy Genetics and Personalized Medicine, Danish Epilepsy Center, Dianalund 4293, Denmark;
    • 11Department of Regional Health Research, Faculty of Health Sciences, University of Southern Denmark, Odense 5230, Denmark;
    • 12Department of Clinical and Experimental Medicine, Pisa University Hospital, Pisa 56126, Italy;
    • 13Division of Paediatric Neurology, Department of Pediatrics, Faculty of Medicine, University of Malaya, Kuala Lumpur 50603, Malaysia;
    • 14Division of Neurology, Department of Medicine, Faculty of Medicine, University of Malaya, Kuala Lumpur 50603, Malaysia;
    • 15Department of Research and Innovation, Division of Clinical Neuroscience, Oslo University Hospital, Oslo 0372, Norway;
    • 16National Centre for Epilepsy, Oslo University Hospital, Oslo 1337, Norway;
    • 17Department of Human Neurosciences, Sapienza University, Rome 00185, Italy;
    • 18Department of Epilepsy Genetics and Personalized Medicine, Danish Epilepsy Center, Dianalund 4293, Denmark;
    • 19Institute of Clinical Medicine, University of Copenhagen, Copenhagen 2200, Denmark;
    • 20Full member of ERN-Epicare, Pediatric Neurology and Muscular Disease Unit, IRCCS Istituto ‘G. Gaslini’, Genoa 16147, Italy;
    • 21Department of Neurosciences, Rehabilitation, Ophthalmology, Genetics, Maternal and Child Health, University of Genova, Genova 16132, Italy;
    • 22Department of Basic and Clinical Neurosciences, Institute of Psychiatry, Psychology and Neuroscience, King's College London, London SE5 9RT, United Kingdom;
    • 23MRC Centre for Neurodevelopmental Disorders, King's College London, London SE1 1UL, United Kingdom;
    • 24King's College Hospital, London SE5 9RS, United Kingdom;
    • 25Genetics and Genome Biology Program, The Hospital for Sick Children, Toronto M5G 0A4, Canada;
    • 26Departments of Statistical Sciences and Computer Science, The University of Toronto, Toronto, M5G 1Z5, Canada
Published June 2, 2026. https://doi.org/10.1101/gr.281480.125
Download PDF Cite Article Permissions Share
cover of Genome Research Vol 36 Issue 6
Current Issue:
June 2026, Vol. 36, No. 6
Focus view

Abstract

The clinical utility of PRS may be hindered by reliance on large, heterogeneous data sets for generation that dilute phenotypic specificity. Meanwhile, small, well-defined clinical cohorts (target) are ubiquitous but insufficient for PRS development. We propose an external-PRS (ePRS) framework borrowing from the transfer learning literature that integrates genetic evidence from target cohorts, incorporating continuous evidence measures and genetic correlation for robust predictions. Simulation indicates superior performance of ePRS across varying genetic correlations between the source and target phenotypes. ePRS refines an idiopathic generalized epilepsy (IGE) PRS to improve differentiation between juvenile myoclonic epilepsy (JME) and other IGE subtypes and, leveraging a large attention deficit hyperactivity disorder GWAS, enhances predictions for impulsivity in JME. Finally, to address concerns about potential cross-platform artifacts, we train and evaluate ePRS in the Canadian Cystic Fibrosis (CF) Gene Modifier Study cohort to predict CF-related diabetes using UK Biobank type 2 diabetes (T2D) summary statistics as the external source phenotype. ePRS continues to improve prediction accuracy in this single-cohort, harmonized-QC setting and offers more precise risk stratification and personalized care across complex traits.

Back to top
Article contents

Announcement(s)