A Genetic Algorithm-Based Clustering Approach for Selecting Non-redundant MicroRNA Markers from Microarray Expression Data

During the last few years, different studies have been done to reveal the involvement of microRNAs (miRNAs) in pathways of different types of cancers. It is evident from the research in this field that miRNA expression profiles help classify cancerous tissue from normal tissue or different subtypes of cancer. In this article, miRNA expression data of different cancer types are analyzed using a novel multiobjective genetic algorithm-based feature selection method for finding reduced non-redundant set of miRNA markers. Three objectives, viz. classification accuracy, a cluster validity index call Davies–Bouldin (DB) index, and the number of miRNAs encoded in a chromosome of genetic algorithm is optimized simultaneously. The classification accuracy is maximized to obtain the most relevant set of miRNAs. DB index is optimized for clustering the miRNAs and choosing representative miRNAs from each cluster in order to obtain a non-redundant set of miRNA markers. Finally, the number of miRNAs is minimized to yield a reduced set of selected miRNAs. The performance of the proposed genetic algorithm-based method is compared with that of the other existing feature selection techniques. It has been found that the performance of the proposed technique is better than that of the other methods with respect to most of the performance metrics. Lastly, the obtained miRNA markers with their associated disease and number of target mRNAs are reported.