Memetic NSGA - a multi-objective genetic algorithm for classification of microarray data

In Gene Expression studies, the identification of gene subsets responsible for classifying available samples to two or more classes is an important task. One major difficulty in identifying these gene subsets is the availability of only a few samples compared to the number of genes in the samples. Here we treat this problem as a Multi-objective optimization problem of minimizing the gene subset size and minimizing the number of misclassified samples. We present a new elitist non-dominated sorting-based genetic algorithm (NSGA) called memetic- NSGA which uses the concept of memes. Memes are a group of genes which have a particular functionality at the phenotype level. We have chosen a 50 gene Leukemia dataset to evaluate our algorithm. A comparative study between Memetic-NSGA and another non-dominated sorting genetic algorithm, called NSGA-II, is presented. Memetic-NSGA is found to perform better in terms of execution time and gene-subset length identified.