Genomic Biomarkers of Metastasis in Breast Cancer Patients: A Machine Learning Approach

Abstract

One of the cancers with the highest incidence in the world is breast cancer (BC). The aim of this study is to identify candidate biomarker genes to predict the risk of distant metastases in patients with BC and to compare the performance of machine learning (ML) based models. In the study; Genomic dataset containing 24,481 gene expression levels of 97 patients with BC was analyzed. Biomarker candidate genes were determined by ML approaches and models were created with XGBoost, naive bayes (NB) and multilayer perceptron (MLP) algorithms. The accuracy values of XGBoost, NB and MLP algorithms were obtained as 0.990, 0.907 and 0.979, respectively. Our results showed that XGBoost has higher performance. The top five genes associated with BC metastasis were AL080059, Ubiquilin 1, CA9, PEX12, and CCN4. In conclusion, when the ML method and genomic technology are used together, the distant metastasis risk of patients with BC can be successfully predicted. The developed XGBoost model can distinguish patients with distant metastases. Identified biomarker candidate genes may contribute to diagnostic, therapeutic and drug development research in patients with metastases.