Background: Breast cancer (BRCA) is a malignant tumor with a high mortality rate and poor prognosis in patients. However, understanding the molecular mechanism of breast cancer is still a challenge.
Materials and methods: In this study, we constructed co-expression networks by weighted gene co-expression network analysis (WGCNA). Gene-expression profiles and clinical data were integrated to detect breast cancer survival modules and the leading genes related to prognostic risk. Finally, we introduced machine learning algorithms to build a predictive model aiming to discover potential key biomarkers.
Results: A total of 42 prognostic modules for breast cancer were identified. The nomogram analysis showed that 42 modules had good risk assessment performance. Compared to clinical characteristics, the risk values carried by genes in these modules could be used to classify the high-risk and low-risk groups of patients. Further, we found that 16 genes with significant differential expressions and obvious bridging effects might be considered biological markers related to breast cancer. Single-nucleotide polymorphisms on the CYP24A1 transcript induced RNA structural heterogeneity, which affects the molecular regulation of BRCA. In addition, we found for the first time that ABHD11-AS1 was significantly highly expressed in breast cancer.
Conclusion: We integrated clinical prognosis information, RNA sequencing data, and drug targets to construct a breast cancer-related risk module. Through bridging effect measurement and machine learning modeling, we evaluated the risk values of the genes in the modules and identified potential biomarkers for breast cancer. The protocol provides new insight into deciphering the molecular mechanism and theoretical basis of BRCA.
Keywords: breast cancer; differential expression; machine learning; single-nucleotide polymorphisms; survival analysis.
Copyright © 2021 Wang, Zhao, Han, Zhang, Zhang, Li and Li.