New paradigms for the design and analysis of oncology clinical trials of molecularly targeted agents are needed. We will cover the targeted “enrichment” design in which a test result is used as an eligibility criterion and the “stratified designs” in which the test is used as part of the primary analysis plan. We will also describe a flexible design for evaluating multiple candidate predictive biomarkers and the recently published cross-validated adaptive signature design for combining in one clinical trial the development and internal validation of a predictive classifier utilizing any type of candidate variables. We will present a framework for the analysis of clinical trials that incorporates both hypothesis testing and predictive modeling but requires that models be justified based on internally validated predictive accuracy.