With the growing understanding of molecular abnormalities in Oncology, there is a shift from histology as driver for treatment selection to molecular target as driver for patient selection. While recent successes are encouraging, larger gains may likely require combination of 2 or more targeted agents. The talk will focus on two specific challenges: (a) the appropriate choice of statistical design and prospective validation of the biomarker for characterization of potential treatment benefit in the targeted subgroup, and (b) development of combination of two or more targeted agents. We will consider prospective validation of the biomarker using an “All Comers” design using Frequentist and Bayesian approaches. For the development of combination of two or more targeted therapies, we consider a Bayesian dose escalation model for the assessment of dose and schedule of the combination drugs.