Abstract
This study systematically screened palmitoylation-related key biomarkers in breast cancer and constructed a high-precision diagnostic model combining multi-omics data and machine learning algorithms. Following batch effect correction across datasets via the ComBat algorithm, 1,782 differentially expressed genes (DEGs) were identified and showed significant enrichment during the extracellular matrix remodeling process and in the PI3K-Akt signaling pathway, highlighting their potential roles in the regulation of cellular microenvironment and the activation of signaling network. A total of 5 core genes (HBB, BGN, CTHRC1, FABP4, CD34) were screened using three machine learning methods. Further analysis of gene contribution revealed that CTHRC1 served as the key gene to affect the prediction of the model. SHAP interpretability analysis highlighted CTHRC1 as a primary contributor to prediction models. Spatial transcriptomic mapping revealed that focal high expression of CTHRC1 in tumor tissues was correlated with enhanced microenvironmental heterogeneity. Functional assays demonstrated that CTHRC1 knockdown can significantly attenuate cell proliferation and clonogenic potential of breast cancer. These findings collectively elucidate that palmitoylated CTHRC1 drives breast cancer progression by enhancing tumor cell proliferation and microenvironment heterogeneity. Clinically, CTHRC1 has shown potential as a diagnostic biomarker and therapeutic target—laying the foundation for the development of liquid biopsies and therapeutic targets based on CTHRC1 to improve the precision management of breast cancer. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12885-025-15190-w.