Integrative machine learning model of RNA modifications predict prognosis and treatment response in patients with breast cancer.

Abstract

Background: Breast cancer, a highly heterogeneous and complex disease, remains the leading cause of cancer-related death among women worldwide. Despite advances in treatment modalities, effective prognostic models and therapeutic strategies are still urgently needed.

Methods: We retrospectively analyzed 15 independent breast cancer cohorts to explore the role of RNA modifications in the prognosis of patients with breast cancer. By integrating nine types of RNA modifications, we developed a comprehensive machine learning-based RNA modification signature (CMRS). Furthermore, single-cell RNA sequencing data were analyzed to understand the biological mechanisms underlying CMRS. In addition, immune infiltration levels were evaluated via six different algorithms, and immune checkpoint inhibitor responsiveness was predicted. Moreover, the response of high-CMIS patients to chemotherapy was predicted via multiple datasets. Finally, immunohistochemistry was performed on tissue samples from breast cancer patients to validate protein expression levels.

Results: Our analysis revealed five key RNA modification-related genes (ENO1, ARAF, WT1, GADD45A, and BIRC3) associated with breast cancer prognosis. The CMRS model demonstrated high predictive accuracy across multiple cohorts and was significantly correlated with patient survival outcomes. Multiomics analysis revealed that high CMRS was associated with increased tumor mutational burden and distinct mutational signatures, particularly in pathways related to TP53, MYC, and cell proliferation. Single-cell analysis highlighted the involvement of epithelial cells and MYC signaling in high CMRS activity. Cell‒cell communication analysis revealed reduced interaction strength in hig CMRS patients, indicating poor prognosis. Furthermore, low CMRS patients presented increased immune cell infiltration and improved responsiveness to immune checkpoint inhibitors, whereas high CMRS patients were identified as potential candidates for treatment with panobinostat and vincristine.

Conclusion: Our study elucidates the significant role of RNA modifications in breast cancer prognosis and treatment. The CMRS model serves as a sensitive biomarker for predicting patient survival and treatment responsiveness, offering a new avenue for personalized therapy in patients with breast cancer.