Developing and evaluating genomics- or proteomics-based diagnostic tests: statistical perspectives

The completion of the Human Genome Project and the ongoing sequencing of mouse, rat, and other genomes have led to an explosion of genetics-related technologies that are finding their way into all areas of biological research in both basic sciences and clinical applications. High-throughput genomics and proteomics technology has been quickly adapted to develop tools for clinical and pharmacological applications. Because molecular alterations usually occur much earlier than histological, physiological, and clinical abnormality, researchers hope to extend the applications of genomics and/or proteomics technology to early diagnosis of diseases and clinical outcome prognosis. Recently, some successful attempts in molecular diagnosis or prognosis have been published. However, for such tests to be translated from the bench to the bed, they must meet some rigorous standards. To develop a clinically meaningful genomics-based diagnostic test, we must have good study design, appropriate statistical analyses, and valid assessment of its clinical efficacy. In this chapter, we discuss statistical considerations on the process of developing reliable and useful genomics- or proteomics-based tests.