Publications

Abstract

Global analysis of gene expression aims to identify key genes, the aberrant expression of which provides insight into the molecular mechanisms of disease. The identification of relevant genes remains challenging, in part because disease heterogeneity ensures that many key genes will not be aberrantly expressed in every specimen. As current gene identification techniques penalise genes with such inconsistent aberrant expression, sometimes to the point of not detecting them at all, there is a need for an analysis methodology with a reduced sensitivity to the frequency of aberrant gene expression across a sample set. Here we present Messina, a method that can identify genes that show more biologically-relevant frequencies of differential expression in disease. We demonstrate with simulated data that Messina is highly sensitive and specific when used for gene selection, compare Messina to contemporary analysis techniques, and validate the aberrant expression of a gene detected by Messina but not by conventional methods. Messina has been packaged into a freely-available stand-alone software program for the analysis of microarray data.