Predicting the Metastasis Ability of Prostate Cancer using Machine Learning Classifiers

Abstract

Patients with prostate cancer (PCA) are more vulnerable to metastasis, which is the disease's most devastating result and the primary reason for mortality. It is still not possible to accurately anticipate whether or not locally advanced PCA will spread. In this work, potential biomarkers are identified by using Machine learning, which compares the gene expressions of metastatic and local prostate cancer by identifying the differentially expressed genes (DEGs) and the molecular pathways associated with the metastasis development of prostate cancer. Two gene profiles (GSE32269 and GSE 6919) are downloaded from the Gene Expression Omnibus collection, which contains a total of 226 tissue samples (69 metastatic, 81 normal prostates, and 76 localized PCA). A fine-tuned Support vector machine(SVM) for feature selection and classification is used, which is employed to analyze gene activity and select vital biomarkers. Moreover,  this study examines the genomic activity and determines the key gene that is essential in distinguishing between localized and metastatic PCA.