Single-cell and spatial transcriptomics reveal post-translational modifications in osteosarcoma progression and tumor microenvironment.

Emerging evidence suggests that post-translational modifications (PTMs) contribute to osteosarcoma pathogenesis, yet their exact molecular roles require further elucidation. Using the AddModuleScore method, we classified tumor cells on the basis of PTMs scores via single-cell RNA sequencing (scRNA-seq). The robust cell type decomposition (RCTD) approach was then applied to map these single-cell groupings onto spatial transcriptomics (ST), enabling the analysis of cell dependencies and the identification of distinct tumor cell subtypes. A prognostic model was constructed using bulk transcriptomic data to predict therapeutic outcomes in immunotherapy. Laboratory experiments were carried out to confirm the biological function of vimentin (VIM). PTMs scores were significantly elevated in tumor cells, stratifying osteoblastic cells (os) into two clusters: PTMs highos and PTMs lowos. The PTMs highos phenotype exhibited pronounced malignant characteristics and closely interacted with fibroblasts in both scRNA-seq and ST analyses. A set of ten hub genes was identified, forming a consensus machine learning-derived post-translational modification gene signature (CMDPTMS) with strong prognostic predictive capability. Although the high-CMDPTMS group (over the median risk score) was linked to poor outcomes and diminished benefit from immunotherapy, seven drugs were identified that may offer therapeutic promise for these patients. Finally, we confirmed that VIM can inhibit the growth and migration of OS cells. In summary, by integrating bulk RNA-seq, scRNA-seq, and ST, we introduced that CMDPTMS may serve as a powerful tool for enhancing OS prognosis prediction and optimizing immunotherapy strategies.