New Bio-Marker Gene Discovery Algorithms for Cancer Gene Expression Profile

Abstract

Several hybrid gene selection algorithms for cancer classification that employ bio-inspired evolutionary wrapper algorithm have been proposed in the literature and show good classification accuracy. In our recent previous work, we proposed a new wrapper gene selection method based-on firefly algorithm named FF-SVM. In this work, we will improve the classification performance of FF-SVM algorithm by proposed a new hybrid gene selection algorithm. Our new biomarker gene discovery algorithm for microarray cancer gene expression analysis that integrates f-score filter method with Firefly feature selection method alongside with SVM classifier named FFF-SVM is proposed. The classification accuracy for the selected gene subset is measured by support vector machine SVM classifier with leave-one-out cross validation LOOCV. The evaluation of the FFF-SVM algorithm done by using five benchmark microarray datasets of binary and multi class. To show result validation of the proposed we compare it with other related state-of-the-art algorithms. The experiment proves that the FFF-SVM outperform other hybrid algorithm in terms of high classification accuracy and low number of selected genes. In addition, we compare the proposed algorithm with previously proposed wrapper-based gene selection algorithm FF-SVM. The result show that the hybrid-based algorithm shoe higher performance than wrapper based. The proposed algorithm is an improvement of our previous proposed algorithm.

Keywords

• Firefly
• microarray
• SVM
• gene selection
• bio-inspired
• gene expression profile