Identification of disease critical Genes for Triple Negative Breast Cancer, TNBC, using Particle Swarm Optimization, PSO, with Machine Learning

Triple Negative Breast Cancer (TNBC) is an aggressive subtype of breast cancer observed in young women. Continuous efforts are being made to identify genes causing Breast Cancer through microarray dataset mining. In this paper, Particle Swarm Optimization has been applied to identify disease causing genes and four classification algorithms, namely, k-Nearest Neighbor (kNN), Support Vector Machine (SVM), Naïve Bayes (NB) and Decision Tree (DT) have been employed to classify the samples. As a result, four hybrid algorithms, namely, PSO-kNN, PSO-SVM, PSO-NB and PSO-DT have been proposed and applied to identify the informative genes for Triple Negative Breast Cancer. Leave One Out (LOO) algorithm is applied for cross validation purpose. Experiments have been conducted on microarray datasets (GSE27447, GSE38959 and GSE31519) to verify the efficiency of algorithms. A comparative study has been performed among the algorithms and biologically relevant genes are identified. Each algorithm has obtained 100% average accuracy in GSE31519 dataset. PSO-DT variant has outperformed all other algorithms in terms of average accuracy in GSE27447 and GSE38959 datasets. Some significant biomarkers identified are: HBB, ITR, GATA, EGFR, ENO2 and ATG9A.