Reproducibility Measures for Microarray Studies

Abstract

Microarrays may be of poor and inconsistent quality for many reasons. Arrays may have no apparent physical imperfections but still yield inconsistent findings relative to other replicate array sets. The reproducibility of microarray results is essential to achieving the benefits of replication and for engendering confidence in scientific conclusions drawn from microarray studies. This paper considers this problem of reproducibility. It opens with a discussion of the issue and its scientific importance and then presents an analytical framework for its study. It considers the data processing steps needed to improve the consistency of replicates, presents statistical measures of reproducibility and demonstrates the ideas and procedures using a case illustration.