Biclustering of DNA microarray data using artificial immune system

Advances in DNA microarray technology has motivated the research community to introduce sophisticated techniques for analyzing the resulted large-scale datasets. Biclustering techniques have been widely adapted for analyzing microarray gene expression data due to its ability to extract local patterns with a subset of genes that are similarly expressed over a subset of samples. Mostly, biclustering methods are based on greedy heuristics which often result in suboptimal solutions. To this end, this paper presents a clonal selection algorithm for biclustering (Bic-CSA) that incorporates these greedy searching procedures as local search heuristics in an immune-inspired algorithm. The quality of biclusters has been demonstrated by experimentation on a well known benchmark dataset. Moreover, the performance of Bic-CSA is compared with other related local search-based methods and immune inspired algorithms. It is shown from results and comparative study that the proposed algorithm outperforms other algorithms in terms of bicluster size and mean-squared residue.