Abstract

Since the different, so-called omics disciplines generate huge amount of data, the application of appropriate, sophisticated statistical methods for developing and validating predictive molecular signatures for drug development, for prevention, screening, diagnosis, monitoring of treatment or aftertreatment of diseases as well as for stratification of individuals is fundamental. The development and validation require several steps and it is quite a long journey from the detection of a molecular predictive signature to the routine use in clinical practice. In our review we focus on data obtained from cDNA expression microarrays. We describe the necessary development and validation steps including recent results of the second phase of the MAQC project (MAQC-II) and emphasise on potential pitfalls.

Keywords: Sample size planing, preprocessing, class prediction, internal validation, external validation