Prioritization of Clusters for Post-genomic Analysis

Clustering algorithms are occasionally used on biological datasets to obtain statistically coherent groups of biomolecules. The biological coherence of such a group (e.g., a gene or protein cluster) might have different interpretations. Functional similarity is the most widely used form of biological coherence of a cluster. We often require to assign priorities, which would signify the betterness of biological coherence, to such clusters for post-genomic analysis. In this paper, we propose a novel approach of prioritizing gene clusters. We introduce a new measure of compactness for quantifying the coherence of the clusters based on their strength of associativity. Employing this, a post processing subroutine is introduced to fit with the standard clustering algorithms to obtain a set of improved prioritized clusters. We test the results upon applying the method on several microarray datasets. Statistical and biological studies on the derived clusters depict the effectiveness of the proposed methodology in the better selection of functionally enriched gene clusters. The proposed methodology helps to select significant clusters and also filter out the noisy ones.