Finding Discriminatory Genes: A Methodology for Validating Microarray Studies

This paper explores the challenge of efficiently collecting data to find which genes (from a given set of candidates) are differentially expressed. We consider several algorithms for this task, including some that assume there are only two types of genes: those that are not differentially expressed, and those that are differentially expressed to the same level. We provide a framework for evaluating such algorithms and also present an algorithm that has nice theoretical properties and performs very well on both real and simulated data.