Hybrid Cuckoo Search with Clonal Selection for Triclustering Gene Expression Data of Breast Cancer

Abstract

Triclustering techniques are applied to analyze three-dimensional gene expression microarray data to retrieve group of genes under the tested samples over certain time points based on a similarity measure. The real-life three-dimensional dataset chosen is estrogen induced breast cancer dataset. In such datasets, identifying the variations in mining the gene expressions across samples and time points is a difficult task. In this work, the hybrid cuckoo search with clonal selection is applied to extract co-expressed genes over samples and times with triclustering solution. The cuckoo eggs are procreated based on the clonal selection theory. Clonal selection generates the variability required for the solution without any substantial modification. The findings are then used to determine the biological significance of genes in the resulting cluster using gene ontology, functional annotation, and transcription factor binding sites. The proposed work identifies the top genes SRY, SOX5, PAX4, NKX61, SP1, OCT, CDP, POU3F2 and PAX4 present in the tricluster which are associated with the breast cancer. To learn the performance of the proposed work, the experiment results are analyzed with conventional cuckoo search. The hybrid cuckoo search with clonal selection outperforms both the conventional cuckoo search and the other existing triclustering algorithms.