Personalized treatment decision-making using a machine learning-derived lactylation signature for breast cancer prognosis.

Background: Breast cancer is a heterogeneous malignancy with complex molecular characteristics, making accurate prognostication and treatment stratification particularly challenging. Emerging evidence suggests that lactylation, a novel post-translational modification, plays a crucial role in tumor progression and immune modulation. Methods: To address breast cancer heterogeneity, we developed a machine learning-derived lactylation signature (MLLS) using lactylation-related genes selected through random survival forest (RSF) and univariate Cox regression analyses. A total of 108 algorithmic combinations were applied across multiple datasets to construct and validate the model. Immune microenvironment characteristics were analyzed using multiple immune infiltration algorithms. Computational drug-repurposing analyses were conducted to identify potential therapeutic agents for high-risk patients. Results: The MLLS effectively stratified patients into low- and high-risk groups with significantly different prognoses. The model demonstrated robust predictive power across multiple cohorts. Immune infiltration analysis revealed that the low-risk group exhibited higher levels of immune checkpoints (e.g., PD-1, PD-L1) and greater infiltration of B cells, CD4 Conclusion: The MLLS represents a promising prognostic biomarker and may support personalized treatment strategies in breast cancer, particularly for identifying candidates who may benefit from immunotherapy.