Individualized selection of chemotherapy regimens based on the molecular characteristics of the cancer would be desirable. However, it is not currently feasible. The development of clinically useful predictors of response to chemotherapy has proven to be difficult. Recently, novel analytical tools particularly transcriptional profiling, have shown promise as potential predictive tests in several small exploratory studies. However, many of the practical challenges of clinical pharmacogenomics are the same that have plagued prognostic and predictive marker research for decades. Perhaps, the most important challenge is to prospectively design and conduct validation trials that demonstrate clinical utility by showing improved patient outcomes with the use of a proposed new test.