Random Forest Algorithm in Unravelling Biomarkers of Breast Cancer Progression

Abstract

Breast cancer is the leading cause of cancer death among women. As its development involves a multidimensional network of gene-environment interactions, advanced data analysis tools and bioinformatics are vital to uncover the nature of cancer. The initial database contained the expression values of 19737 genes in 1082 patients. Random Forest algorithm was used to distil the genes with the strongest influence on four substantial prognostic factors (survival period, tumour size, lymph node seizure, and metastasis). The obtained set consists of 230 potential biomarkers that facilitate the critical cancer-related pathways, such as p53, Wnt, VEGF, UPP, thereby influencing cell proliferation, tumouri- and angiogenesis. A considerable contrast in the expression was shown between the patients at different stages of cancer progression. The obtained set will simplify the diagnostics and prediction of tumour progression, enhance treatment outcomes and elaborate better strategies for curing breast cancer.