Identifying breast cancer recurrence risk factors via Bayesian elastic net quantile regression

This study investigates the use of Bayesian Elastic Net Quantile Regression (BENQReg) for identifying recurrence risk factors in breast cancer patients. Unlike traditional regression models that estimate average effects, BENQReg captures how predictor influence varies across different quantiles of the recurrence-free survival distribution. The model integrates an asymmetric Laplace likelihood with elastic net regularization in a hierarchical Bayesian framework, enabling both variable selection and quantile-specific inference.A comprehensive simulation study was conducted to evaluate the model’s performance under conditions of sparsity and predictor correlation. The results showed high estimation accuracy, perfect true positive rates, and robust behavior across quantile levels. BENQReg consistently outperformed standard Bayesian quantile models in variable recovery, especially at the median quantile level.The model was then applied to a breast cancer dataset containing clinical, biomarker, and genetic features. BENQReg identified distinct predictors at different quantiles, including tumor size and HER2 status as early risk indicators, and hormone receptor status and BRCA2 expression as protective factors in patients with longer recurrence-free survival. These findings demonstrate the model’s capacity to detect quantile-dependent associations, offering clinically interpretable insights that support precision medicine.BENQReg provides a robust, flexible, and informative approach for analyzing heterogeneous biomedical data and is especially valuable in high-dimensional settings where variable selection is critical.