Integrated transcriptomics identifies a high-glycolysis osteosarcoma phenotype and links SSR3 to stemness and chemoresistance

Osteosarcoma treatment is hampered by intratumoral metabolic heterogeneity. We analyzed single‑cell and spatial transcriptomic data from 27 patients and stratified malignant cells into high- (HGlyOS), medium-, and low-glycolysis groups. HGlyOS cells exhibited enhanced stemness and chemotherapy resistance features and occupied an early inferred transcriptional state. Using nested machine learning across datasets, we derived a 34-gene Glycolysis activity Prediction System (GGPS) as a parsimonious proxy for the Composite Glycolytic Activity Score (CGAS)-defined glycolytic phenotype rather than as an independent subtype classifier. Through a separate screening pipeline combining co‑expression network analysis, differential expression, and machine learning, followed by literature mining, we identified SSR3 (which was not part of the 34‑gene signature) as a previously unreported glycolysis regulator in osteosarcoma. Independent functional experiments confirmed that SSR3 modulates glycolytic metabolism, stemness maintenance, and cisplatin sensitivity. Our work provides a metabolic classification framework for osteosarcoma and a proof‑of‑concept tool for single‑cell glycolytic phenotyping.