Background: Breast cancer, as one of the most common malignant tumors among women worldwide, its heterogeneity, invasiveness and drug resistance pose great challenges to clinical treatment. Lactylation, as an emerging post-translational modification of proteins, has been proven to play an important role in tumor biology, especially in the regulation of the tumor microenvironment (TME) and immune microenvironment. However, the role of lactylation in breast cancer is not yet fully understood. Methods: Download the RNA-seq data and clinical data of breast cancer from TCGA, GEO and Metabric databases. Use Cox regression analysis and LASSO analysis to build a risk prediction model, and use Kaplan - Meier survival analysis and ROC curve to verify the accuracy of the model. In addition, use the consensus clustering method to determine the relevant clusters of breast cancer patients and verify the clinical predictive significance of the lactylation model genes again. We also conduct a detailed exploration of the potential mechanisms by which the lactylation model genes lead to poor prognosis in patients through single-cell and spatial transcriptomics (ST). Results: Six lactylation-related breast cancer prognosis models with prognostic value for patients were constructed, and these six genes were all risk factors for breast cancer patients. Through the consensus clustering method, it was found that these six lactylation-related DEGs (LRDEGs) could also be used to classify breast cancer patients, and the subtypes with poor prognosis had high expression of these genes. Then, through immune analysis, obvious immune heterogeneity was found among patients of different subtypes. In order to explore the potential mechanism of poor prognosis, single-cell and ST analysis revealed that CORO6+Epi might be the lactylation-related subtypes leading to poor prognosis in breast cancer. SCENIC transcription factor analysis identified that E2F3 might drive the formation of the malignant phenotype of CORO6+Epi. Finally, through cell communication analysis, it was found that CORO6+Epi might be regulated by POSTN fibroblasts through the THBS2 - SDC4 axis to promote tumor progression. Conclusion: A new lactylation-related gene prediction model was successfully constructed, which can accurately predict breast cancer patients with poor prognosis. It was also found that this might be related to the transcription factor E2F3 and the THBS2 - SDC4 axis between CORO6+Epi and POSTN fibroblasts, providing new therapeutic targets for the precision treatment of breast cancer.