Integrating quality-based clustering of microarray data with Gibbs sampling for the discovery of regulatory motifs

In microarray experiments, genes exhibiting a similar expression profile are potentially coregulated. Clustering identifies such groups of coexpressed genes, whose upstream regions can then searched for putative regulatory elements. We present two algorithms and an interactive web-based user interface that integrate cluster analysis and motif finding for the analysis of microarray data. Starting from the expression, we present our adaptive quality-based clustering algorithm to define groups of tightly coexpressed genes. The upstream region is then retrieved based on the accession number and gene name. Once the upstream regions are identified, the sequences are analyzed using Gibbs sampling for motif finding to find the over-represented motifs. Our implementation (called Motif Sampler) allows the use of higher-order models for the sequence background. This methodology can be used through our INCLUSive web interface at the following URL: http://www.esat.kuleuven.ac.be/~dna/BioI/Software.html