Introduction to prediction modeling using machine learning and omics data

Prediction models for prevention, diagnosis, prognosis, and treatment selection are an important component of precision medicine. However, there are many methodological challenges in developing scientifically valid and meaningful prediction models based on clinical or molecular data. This chapter reviews commonly used terminology, the general workflow for developing and assessing prediction models, and provides a brief introduction to selected machine learning (ML) approaches. It also presents measures of prediction performance, describes the crucial step of internal and external validation, and introduces methods to improve the interpretability of prediction models. The aim of this chapter is to provide clinicians and medical scientists with a basic understanding of prediction modeling using ML approaches. It will enable them to critically evaluate the literature and to assist statisticians and computer scientists in developing clinically relevant prediction models.