A Bayesian precision medicine framework for calibrating individualized therapeutic indices in cancer

The development and clinical implementation of evidence-based precision medicine strategies has become a realistic possibility, primarily due to the rapid accumulation large-scale genomics pharmacological data from diverse model systems: patients, cell lines drug perturbation studies. We introduce novel Bayesian modeling framework called individualized theRapeutic index (iRx) integrate high-throughput pharmacogenomic across systems. Our iRx achieves three main goals: first, it exploits conserved biology between patients calibrate therapeutic response drugs in patients; second, finds optimal line avatars as proxies for patient(s); finally, identifies key genomic drivers explaining line-patient similarities. This is achieved through semi-supervised learning approach that conflates (unsupervised) sparse latent factor models with (supervised) penalized regression techniques. propose unified tractable estimation, inference conducted via efficient posterior sampling schemes. illustrate validate our using two existing trial sets multiple myeloma breast cancer show improves prediction accuracy compared naive alternative approaches, consistently outperforms methods literature both simulation scenarios well real examples.