ABSTRACT
In the field of predictive genomic studies, two main approaches are applied to the problem of class identification: class discovery, and class prediction. This chapter focuses on class prediction methods in clinical studies, reviews the distinction between these two approaches. It attempts to lay a basic foundation for the principles of class prediction in clinical pharmacogenomic studies and how questions regarding toxicity can be answered during the course of these studies. The chapter highlights many of the issues facing scientists attempting to perform clinical pharmacogenomic analyses, and introduces a general approach to clinical pharmacogenomic studies conducted in concert with clinical trials. The most important initial goal and potential benefit of toxicogenomics will be to accelerate the evaluation of toxicity for new chemical entities. Biomarkers measured at early time points that are significantly correlated with either eventual response or eventual toxicity would be of obvious importance in monitoring the progress of patients prior to the end of therapy.
