Development of a single-cell derived MDSCs signature score for prognostic risk stratification and therapeutic decision guidance in breast cancer

• Novel MDSCs Gene Signature: First systematic identification of 209 breast cancer-specific MDSCs signature genes via integrated single-cell RNA-seq and bulk multi-omics analysis, providing a molecular basis for MDSCs characterization in the tumor microenvironment (TME). • Robust Prognostic Model: Development of a 5-gene (BCL2A1, GDI2, GRINA, RNASE1, SERPINA1) risk score model with strong prognostic value, achieving 1–10 year overall survival AUC > 0.6 and independent predictive capacity across TCGA-BRCA, METABRIC, and SCANB cohorts. • TME and Therapy Response Stratification: High-risk patients exhibit an immunosuppressive TME (enriched M2 macrophages/regulatory T cells, reduced cytotoxic T lymphocyte activity) and poor responses to chemotherapy/immunotherapy, while low-risk patients show heightened drug sensitivity and favorable immunotherapeutic outcomes. • Clinically Translatable Tools: Construction of a nomogram integrating the MDSCs risk score with clinicopathological factors (age, TNM stage) to enhance personalized survival prediction, validated by calibration plots and decision curve analysis. • Precision Oncology Implications: The risk score system enables tri-dimensional clinical guidance—prognostic stratification, chemotherapy sensitivity assessment, and immunotherapy response prediction—supporting MDSCs-targeted therapeutic strategies in breast cancer. Myeloid-derived suppressor cells (MDSCs) function as critical immunosuppressive constituents within the breast cancer tumor microenvironment (TME). However, their molecular variability and clinical translation potential remain inadequately characterized. By combining single-cell RNA sequencing (scRNA-seq) with bulk multi-omics datasets, we screened and confirmed MDSCs signature genes specific to breast cancer. A prognostic risk scoring model was constructed using machine learning approaches and tested across multiple independent patient cohorts. The model’s predictive capacity for chemotherapy sensitivity, immunotherapy responsiveness, and TME features was systematically assessed. After integrating the single-cell datasets GSE161529 and GSE176078, we identified 12,767 MDSCs along with their 209 characteristic genes. To ensure the reliability of our results, we employed various analytical methods and utilized diverse data for validation. From this signature, a 5-gene risk score model comprising BCL2A1, GDI2, GRINA, RNASE1, and SERPINA1 was constructed, demonstrating robust prognostic stratification with a 1- to 10-year overall survival (OS) AUC exceeding 0.6 and independent predictive value. High-risk patients exhibited characteristic features of an immunosuppressive tumor microenvironment (TME), including increased M2 macrophages and regulatory T cells, alongside diminished cytotoxic T lymphocyte activity. These patients also showed poor responses to both chemotherapy and immunotherapy. This investigation marks the initial systematic characterization of a new MDSCs gene signature in breast cancer, alongside the establishment of an MDSCs-associated marker scoring framework with multi-aspect clinical translation capability, thereby linking MDSCs fundamental biology to precision oncology in this cancer type.