| Title: | Polygenic Stroke Risk and Autoimmune Disease: Associations and Clinical Modifiers |
| Journal: | IEEE Transactions on Biomedical Engineering |
| Published: | 9 Jun 2026 |
| Pubmed: | https://pubmed.ncbi.nlm.nih.gov/42262940/ |
| DOI: | https://doi.org/10.1109/tbme.2026.3702042 |
Publication 20754
WARNING: the interactive features of this website use CSS3, which your browser does not support. To use the full features of this website, please update your browser.
Abstract
Polygenic risk scores (PRSs) can improve prediction of complex diseases, but their utility may vary across clinical subgroups with distinct inflammatory or immunologic profiles. Because autoimmune diseases involve chronic immune dysregulation, they may influence both baseline stroke risk and the added value of genetic information in risk prediction. Using the UK Biobank, we identified individuals with prevalent autoimmune diseases at baseline and evaluated whether PRSs differentially improve 10-year incident stroke prediction in those with versus without autoimmune disease. We modeled 10-year incident stroke using logistic regression, gradient boosting, and XGBoost with demographic, ancestry, and clinical covariates, with and without PRSs. To quantify individual-level PRS contribution, we compared prediction error between non-genetic (NG)-only and NG+PRS models, defining subgroups in which predictions improved or worsened after inclusion of PRSs. We further examined enrichment of autoimmune conditions using approximately 3,600 clinical features across these subgroups. We find that elevated baseline stroke risk in individuals with autoimmune disease reduces the incremental predictive benefit of PRSs, consistent with established evidence linking immune-mediated disease to cerebrovascular risk. By integrating large-scale clinical phenotypes, genetic risk scores, autoimmune status, and longitudinal outcomes within a unified modeling framework, this study reproduces known clinical relationships and quantifies interactions between genetic and immune-mediated risk. Overall, these results clarify when genetic risk adds meaningful predictive value beyond strong clinical signals and highlight a scalable approach for identifying context-dependent utility of PRSs.</p>
8 Authors
- Andrew Hornback
- Yifei Wang
- Marco Piazzi
- Harinishree Sathu
- Kyungbeom Kim
- Pavithra Avula
- Monica Isgut
- May D. Wang
Enabling scientific discoveries that improve human health