Gene expression profiling in cancer research

Gene expression profiling is increasingly used in cancer research. For each patient, the expression of thousands of genes in the tumour can be measured simultaneously on a microarray. Microarray studies aim at classifying patients based on two types of classification schemes: unsupervised classification, which uses clustering in order to identify homogeneous subtypes of a disease on the basis of gene expression, or supervised classification, which principally aims at the identification of genes or set of genes differentially expressed between tumours with different characteristics (molecular signature), for instance between a group of patients with bad and good prognosis. The data consists of a small number of patients and a large number of variables, raising serious methodological problems. We will use published results on breast cancer in order both to study the power of the experiments and to illustrate the problems in interpretation and validity of their results. We recommend rigorous evaluation of this new technology.