Semi-Parametric Bayesian Additive Regression Trees for Risk Prediction with High-Dimensional Epigenetic Signatures and Low-Dimensional Covariates
Researchers have developed a new semi-parametric Bayesian Additive Regression Trees (spBART) model to improve risk prediction using high-dimensional epigenetic data alongside lower-dimensional covariates. This method separates the modeling of low-dimensional covariates into a parametric component for interpretability and uses tree ensembles for complex, high-dimensional predictors. Applied to multiple myeloma studies, the spBART model successfully identified key genetic loci and achieved a strong out-of-sample discrimination AUC of 0.96. AI
IMPACT Introduces a novel statistical framework for integrating complex biological data, potentially advancing precision medicine and risk assessment in disease.