Bayesian Nonparametric Predictive Modeling for Personalized Treatment Selection

Pedone, Matteo Argiento, Raffaele Stingo, C.We develop a Bayesian nonparametric predictive model to establish personalized therapeutic strategies for oncology patients. We leverage characteristics of both the patient and disease to support decision making in the selection of the optimal treatment. The core component of the model is a product partition model with covariates (ppmx) that induces clusters of observations that are more homogeneous with respect to predictive biomarkers. We conduct a simulation study to evaluate different modeling choices regarding ppmx in the framework of personalized treatment selection.