Development and Validation of a SHAP-Interpretable Machine Learning Model for Stroke Risk Prediction Using Circulating MicroRNA Biomarkers
A machine-learning model using blood RNA signatures predicts stroke risk with interpretable results, though larger external validation is needed.
A SHAP-interpretable SVM model using serum miRNA biomarkers from 1,785 samples was trained and externally validated for stroke risk prediction, achieving near-perfect training accuracy and 80% cross-platform external validation accuracy. While the perfect training AUC and tiny external validation cohort (n=10) limit generalizability, the miRNA signature and interpretable model framework are of methodological interest.
What the study was
- Study design
- Machine learning model development and external validation study
- Population
- Stroke patients and non-stroke controls (n=1785 serum miRNA profiles); external NGS validation cohort n=10
- Sample size
- 1785
- Category
- Diagnostics
- Maturity
- Exploratory
- Journal
- Journal of Molecular Neuroscience
Why it surfaced
Large training set (n=1785) with SHAP interpretability; external validation n=10 is extremely small. Perfect training AUC (1.0) suggests overfitting concern. Classification_confidence medium due to validation sample limitations.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.