Studying combined breast cancer biomarkers using machine learning techniques

Abstract

The spread of breast cancer and its high fatality has spurred a lot of research for studying its causes and treatments. Since the discovery of gene extraction methods, many biomarkers have been investigated and related to cancer. The large number of genes and their intertwining relations necessitates advanced machine learning models, rather than simple statistical and correlation analysis. Having the goal to advance the current state of knowledge concerning early diagnosis of breast cancer, we used decision trees, random forest, K-nearest neighbor, SVM, and Gaussian process classifiers, combined with testing different and novel biomarkers. The study showed that the LAPTM4B expression level is more indicative than its counter alleles. Moreover, the combination of biomarkers and machine learning led to enhancement in accuracy over single marker with at least 10%. By measuring the markers' importance, we found that LAPTM4B and OPG combined with age has shown a significant increase in the diagnosis accuracy.