Breast cancer (BC) is the malignancy with the highest mortality rate among women, identification of immune-related biomarkers facilitates precise diagnosis and improvement of the survival rate in early-stage BC patients. 38 hub genes significantly positively correlated with tumor grade were identified based on weighted gene coexpression network analysis (WGCNA) by integrating the clinical traits and transcriptome analysis. Six candidate genes were screened from 38 hub genes basing on least absolute shrinkage and selection operator (LASSO)-Cox and random forest. Four upregulated genes (CDC20, CDCA5, TTK and UBE2C) were identified as biomarkers with the log-rank p < 0.05, in which high expression levels of them showed a poor overall survival (OS) and recurrence-free survival (RFS). A risk model was finally constructed using LASSO-Cox regression coefficients and it possessed superior capability to identify high risk patients and predict OS (p < 0.0001, AUC at 1-, 3- and 5-years are 0.81, 0.73 and 0.79, respectively). Decision curve analysis demonstrated risk score was the best prognostic predictor, and low risk represented a longer survival time and lower tumor grade. Importantly, multiple immune cell types and immunotherapy targets were observed increase in expression levels in high-risk group, most of which were significantly correlated with four genes. In summary, the immune-related biomarkers could accurately predict the prognosis and character the immune responses in BC patients. In addition, the risk model is conducive to the tiered diagnosis and treatment of BC patients.
Keywords: Breast cancer; Immune cell infiltration; Immunotherapy targets; Machine learning; Risk score.
© 2023. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.