Identification of Prognostic and Heterogeneous Breast Cancer Biomarkers Based on Fusion Network and Multiple Scoring Strategies

Abstract

Breast cancer is a malignant disease that is caused by multiple factors, and the prognosis of breast cancer patients is the focus of medical research. In the present study, we have proposed a novel computational framework which identifies the prognostic biomarkers of breast cancer based on multiple network fusion and multiple scoring strategies. In order to eliminate the heterogeneity of samples, we first clustered the patient samples according to the principle components of gene expression. For each cluster, we used the fusion network to reduce the incompleteness of interactome and to take into account more disease-related information. Genes were weighted from the perspectives of biological functions, prognostic ability and correlation with known disease genes, and a network propagation model was applied to the fusion network so as to comprehensively evaluate the influence of genes on breast cancer patients. To evaluate the performance of our method, we have compared our approach with three state-of-the-art approaches. The results demonstrated that biomarkers captured by our method have both strong discriminative power in differentiating patients with different prognostic outcomes and biological significance.