Clustering gene expression data with hybrid GA approach

The combined interpretation of gene expression data and gene sequences offers a valuable approach to investigate the intricate relationships involving gene transcriptional regulation. The highly interactive gene expression data produced by microarray hybridization experiments allow us to find coexpressed genes. By analyzing the upstream regions of the identified coexpressed genes, we can discover the regulatory patterns characterized by transcription factor binding sites, which govern the process of transcriptional regulation. In the following, we present a generic clustering algorithm that uses a Hybrid GA approach to discover clusters in gene expression data. The advantage of this method is that large search space can be effectively explored by utilizing the evolutionary algorithm techniques. Moreover, it is able to discover underlying patterns in noisy gene expression data for meaningful data groupings, and statistically significant patterns hidden in each cluster can also be extracted at the same time. Since, the proposed method can handle both continuous- and discrete-valued data, it can be used with other microarray data and biomedical data. The experimental results obtained from real expression data reveal meaningful groupings and uncover many known transcription factor binding sites.