Microarray data mining: a novel optimization-based approach to uncover biologically coherent structures.

BACKGROUND: DNA microarray technology allows for the measurement of genome-wide expression patterns. Within the resultant mass of data lies the problem of analyzing and presenting information on this genomic scale, and a first step towards the rapid and comprehensive interpretation of this data is gene clustering with respect to the expression patterns. Classifying genes into clusters can lead to interesting biological insights. In this study, we describe an iterative clustering approach to uncover biologically coherent structures from DNA microarray data based on a novel clustering algorithm EP_GOS_Clust. RESULTS: We apply our proposed iterative algorithm to three sets of experimental DNA microarray data from experiments with the yeast Saccharomyces cerevisiae and show that the proposed iterative approach improves biological coherence. Comparison with other clustering techniques suggests that our iterative algorithm provides superior performance with regard to biological coherence. An important consequence of our approach is that an increasing proportion of genes find membership in clusters of high biological coherence and that the average cluster specificity improves. CONCLUSION: The results from these clustering experiments provide a robust basis for extracting motifs and trans-acting factors that determine particular patterns of expression. In addition, the biological coherence of the clusters is iteratively assessed independently of the clustering. Thus, this method will not be severely impacted by functional annotations that are missing, inaccurate, or sparse.

Evaluation of normalization and pre-clustering issues in a novel clustering approach: global optimum search with enhanced positioning.

We study the effects on clustering quality by different normalization and pre-clustering techniques for a novel mixed-integer nonlinear optimization-based clustering algorithm, the Global Optimum Search with Enhanced Positioning (EP_GOS_Clust). These are important issues to be addressed. DNA microarray experiments are informative tools to elucidate gene regulatory networks. But in order for gene expression levels to be comparable across microarrays, normalization procedures have to be properly undertaken. The aim of pre-clustering is to use an adequate amount of discriminatory characteristics to form rough information profiles, so that data with similar features can be pre-grouped together and outliers deemed insignificant to the clustering process can be removed. Using experimental DNA microarray data from the yeast Saccharomyces Cerevisiae, we study the merits of pre-clustering genes based on distance/correlation comparisons and symbolic representations such as {+, o, -}. As a performance metric, we look at the intra- and inter-cluster error sums, two generic but intuitive measures of clustering quality. We also use publicly available Gene Ontology resources to assess the clusters' level of biological coherence. Our analysis indicates a significant effect by normalization and pre-clustering methods on the clustering results. Hence, the outcome of this study has significance in fine-tuning the EP_GOS_Clust clustering approach.