Machine Learning for Precision Medicine
Xin Huang, Yi‐Lin Chiu · 2022
The identification of prognostic and predictive biomarkers is an important scientific component in advancing the drug discovery and development pipeline and closely aligns with the precision medicine definition on disease prognosis prediction and treatment response prediction. With multiple candidate biomarkers available, one can also develop machine learning (ML) algorithms to combine multiple biomarkers into a single signature. Most of regression types of statistical models and ML methods are directly applicable. The actual methods for prognostic marker discovery depend on the study-specific goals and hypothesis. The chapter focuses on the discussion of how to modify popular ensemble learning methods to identify important predictive biomarkers and discusses the topic of subgroup identification for treatment responders. Receiver operating characteristic-based cutoff determination is an important technique, and it has been widely used for not only subgroup identification but also assay and diagnostic test development. The chapter introduces a general framework for the cutoff selection process, called Bootstrapping and Aggregating of Thresholds from Trees.