Novel Hybrid PSO-SA Model for Biclustering of Expression Data

AbstractUncovering genetic pathways is equivalent to finding clusters of genes with expression levels that evolve coherently under subsets of conditions. Conventional clustering approaches like K-Means, heuristic biclustering approaches like Greedy Algorithms, meta-heuristic biclustering approaches like Genetic Algorithm, Simulated Annealing or Particle Swarm Optimization may not be specific enough to find correlations between genes in an appropriate manner, as each of the algorithms has its own merits and demerits. In this paper, a hybrid PSO-SA BIClustering algorithm namely PSO-SA-BIC that combines features of binary PSO with Simulated Annealing has been proposed in order to extract biclusters of gene expression data. A novel fitness function, based on ACV has been used to identify shifting pattern and scaling pattern biclusters. Experimental results on bench mark datasets show that PSO-SA-BIC algorithm is outperforming the classical algorithms by providing statistically significant bicluster