An MRMR with mean score feature selection for ovarian cancer classification using joint analysis

Abstract

Cancer Classification from microarray expression profiles is a challenging task due to its high dimensionality in the field of biomedicine and bioinformatics. The microarray data experiment contains large number of features and small number of samples, therefore feature selection is an essential task in cancer classification. In this paper,  a novel feature selection technique is proposed based on minimum Redundancy Maximum Relevance (mRMR) in which the mean score is introduced to improve the relevance between features. The feature selection is employed in gene expression data and miRNA expression data using joint analysis in ovarian cancer dataset. Joint analysis gives 100% accuracy for ovarian cancer using the classifiers Support Vector Machine (SVM) and Artificial Neural Networks (ANN). The identified signature of miRNAs and genes are useful for finding the stages of ovarian cancer and therapeutic leads to the cancer patients.