Clustering replicated microarray data via mixtures of random effects models for various covariance structures

摘要

A unified approach of mixed-effects model has been recently proposed for clustering correlated genes from different kinds of microarray experiments. With the so-called EM-based MIXture analysis WIth Random Effects (EMMIX-WIRE) model, both the gene-specific and tissue-specific random effects are taken into account in the (mixture) modelling of microarray data. In this paper, we focus on the applications of the EMMIX-WIRE model to the cluster analysis of microarray data with repeated measurements. In particular, we investigate various forms of covariance structure commonly applicable for replicated microarray data and compare their impact on the final clustering results, using a real data set of microRNA profile and a published yeast galactose data set with known Gene Ontology (GO) listings.