Machine learning-based programmed cell death gene signature for prognosis and drug sensitivity in breast cancer.

Breast cancer (BRCA) heterogeneity necessitates robust prognostic biomarkers. Programmed cell death (PCD) serves a key role in tumor progression and therapy response. However, the prognostic potential of PCD-related genes (CRGs) in BRCA remains to be fully elucidated. Therefore, the present study integrated transcriptomic data from The Cancer Genome Atlas, Molecular Taxonomy of Breast Cancer International Consortium and Gene Expression Omnibus databases. Differentially-expressed CRGs were identified in tumor tissues and subjected to univariate Cox regression analysis. A comprehensive machine learning framework, encompassing 101 algorithm combinations, was applied to construct an optimal PCD-based gene signature (CDS). The prognostic value of the CDS, and its association with the tumor immune microenvironment (TIME), predictive power for immunotherapy and drug sensitivity were systematically evaluated using the 'immunedeconv' and 'OncoPredict' R packages. A five-gene CDS (anoctamin 6, polo-like kinase 1, solute carrier family 7 member 5, tubulin α-1C chain and transcobalamin 1) was developed using the Stepwise Cox (both) + Elastic Net (α=0.9) model, demonstrating notably increased predictive performance (concordance index=0.79). High CDS scores were found to be independent prognostic factors for inferior overall survival and were associated with an immunosuppressive TIME, characterized by reduced CD8 T-cell infiltration and impaired immune activation. By contrast, low CDS scores predicted favorable responses to immune checkpoint inhibitors in three immunotherapy cohorts and increased sensitivity to chemotherapy, endocrine and targeted agents. Human Protein Atlas and reverse transcription-quantitative PCR validation were consistent with the dysregulation of CDS genes in BRCA tissues and cell lines. In conclusion, the present study developed and validated a novel, machine learning-derived CDS that effectively stratified BRCA risk in patients. This signature provides insights into the immune landscape and serves as a valuable biomarker in predicting prognosis, immunotherapy efficacy and drug sensitivity, facilitating personalized treatment strategies in the future.