Development and validation of an interpretable machine learning model identify the lactylation-related protein SUSD3 as a prognostic and therapeutic biomarker for breast cancer.

Background: Breast cancer is one of the most prevalent malignancies and a leading cause of cancer-related mortality among women. Lactylation, a recently recognized post-translational modification, has emerged as a significant factor in tumor biology, with increasing evidence linking it to cancer progression and immune modulation. However, the role of lactylation in tumorigenesis remains ambiguous. This raises questions about whether it serves as a primary driver or a secondary regulator during cancer development, as well as its influence on the tumor immune microenvironment and prognostic implications. Methods: This study investigates the clinical relevance of lactylation-related genes (LRGs) in breast cancer through a comprehensive analysis of extensive genomic datasets, including single-cell RNA sequencing, bulk transcriptomic data, and spatial transcriptomics from established public databases such as TISCH, TCGA, and GEO. Results: By using a combination of multiple machine-learning algorithms, we developed an effective lactylation-related signature that correlates with immune cell infiltration, chemokine expression, and tumor mutation burden. This signature proved useful in identifying breast cancer patients likely to respond to immunotherapy. Finally, we experimentally validated the quantified expression levels of hub genes in human breast samples and demonstrated the role of SUSD3. Conclusion: These findings indicate that our lactylation risk model can be used to predict the malignant progression and immune evasion of breast cancer. It is expected to become a potential therapeutic target and a diagnostic marker for breast cancer. This model also provides insights into breast cancer therapy and an effective framework for developing gene screening models applicable to other diseases and pathogenic mechanisms.