Looking for structure in gene expression profiles

A. CHERNOMORETZ

Encuentro; 46 Annual Meeting Argentine Society for Biochemistry and Molecular Biology; 2010

Argentine Society for Biochemistry and Molecular Biology

Typical DNA microarray assays produce a huge amount ofinformation, and clustering techniques are usually employed toreveal common patterns of gene expression across differentsamples. The rationale behind this approach is a ´guilty-byassociation´scenario, where genes with similar profiles of activityare supposed to have related functions or to be regulated bycommon mechanisms. In this communication we introduce avariant of a clustering procedure rooted in statistical mechanicsparadigms: the Super Paramagnetic Clustering. The proposednovel algorithm seeks to find biologically meaningful geneclusters, integrating microarray transcriptional data withbiological information about gene functions, as provided by theGene Ontology database. The new algorithm was applied to severalgene expression data sets. We found out that it greatly improves thebiological coherence of cluster partitions, compared to the naiveSPC implementation, and other traditional clustering methods. Inconclusion, using a combined metric, we present a method that cansimultaneously mine for meaningful structure in both, expressionand functional spaces. The algorithm shows a robust behavior andsucceeds in recognize sensible gene clusters that might be used toidentify relevant biological process for the phenotypes of interest.