Exploring a specialized lactylation pattern to predict the prognosis and sensitivity of immunotherapy in breast cancer via machine learning.

BACKGROUND: With the global incidence of breast cancer steadily increasing, there is a critical need to explore novel molecular mechanisms that drive tumor progression and affect therapeutic response. Lactylation, a recently discovered post-translational modification, has been implicated in various aspects of tumor biology; however, its specific role in breast cancer (BRCA) remains poorly understood. Understanding the association between lactylation and BRCA may offer new insights into prognosis and precision treatment strategies. This study proposes a new framework for understanding the interplay between lactylation and breast cancer progression, with potential implications for prognostic modeling and immunotherapy guidance. METHODS: We performed a comprehensive integrative analysis using multi-omics data from TCGA-BRCA and three independent GEO cohorts. Breast cancer patients were classified into three distinct clusters based on the expression of lactylation-related genes, each exhibiting unique clinical and molecular profiles. To construct a robust prognostic model, 79 machine learning algorithms were evaluated, and the random survival forest (RSF) method was selected as the optimal approach. PGK1 was identified as the most critical predictive gene in the model. RESULTS: The RSF-based risk score demonstrated strong prognostic power and was significantly associated with immune cell infiltration and response to immunotherapy in BRCA patients. Functional validation experiments further confirmed that PGK1 knockdown (shPGK1) markedly suppressed breast cancer cell proliferation and migration in vitro. CONCLUSIONS: We developed a novel lactylation-related prognostic model that effectively predicts immunotherapy outcomes in breast cancer and highlights the central role of PGK1 in modulating tumor behavior and immune response.