Biclustering of Gene Expression Data Based on Binary Artificial Fish Swarm Algorithm

Abstract

Many existing biclustering algorithms has been used to determine co-expressed genes in gene expression data under subsets of experimental conditions. The Mean Squared Residue (MSR) or the Average Correlation Value (ACV) often be employed as fitness functions. But this measure may not find some relevant genes with shifting and scaling patterns. Here we introduce a new approach - Binary Artificial Fish Swarm Algorithm (BAFSA), which possesses an improved Meta-heuristic search algorithm that combines traditional artificial fish swarm algorithm (AFSA) with binary forms. To find genes with shifting and scaling patterns, we used a fitness function based on the linear correlation. The biclustering algorithm based on BAFSA has been applied to Mice Protein Expression dataset and many biologically significant biclusters are found, which exhibited the superb performance. Then the performance of the proposed method is compared to CC, QUBIC and FLOC.