Using multi-objective evolutionary computing for biclustering of gene expression data

GALLO CRISTIAN; CARBALLIDO JESSICA ANDREA; PONZONI IGNACIO

Buenos Aires, Argentina

Workshop; VI ALIO/EURO 2008 (Workshop on Applied Combinatorial Optimization); 2008

A new memetic algorithm that combines a Multi-Objective Evolutionaty Algorithm (MOEA)with a local search strategy for microarray data analysis is presented in this article. The method guides the search through the gene expression data matrix in order to find biclusters that fulfill several objectives. The MOEA used for this aim is based on the SPEA2. The hybrid strategy was compared against a well known method presented in the literature. The case study was a data set corresponding to the Saccharomyces cerevisiae organism. Our algorithm outperformed the previous method in terms of two metrics, namely Set Coverage and Spacing. We are aware that more experiments with data from other organisms are necessary, thus leading to more concluding results. None-theless, the outcomes obtained so far are undeniably promising.