Single-cell and bulk transcriptomic integration identifies a four-gene prognostic score linked to immunometabolic features in osteosarcoma

Background Recurrent and metastatic osteosarcoma calls for prognostic markers that connect tumor-cell states with the metabolic and microenvironmental context of disease progression. We investigated whether recurrence-derived single-cell signals could identify an outcome-associated transcriptional program and functionally informative candidate genes. Methods Single-cell data from GSE152048 were integrated with the 84-patient TARGET-OS cohort to derive a four-gene RiskScore and characterize molecular subtypes, pathway activity, and immune/stromal associations. COL13A1- and KIF25-directed perturbations were examined using transcript and protein measurements, colony formation, and wound-closure assays in 143B and HOS cells. Results Oxidative phosphorylation enrichment characterized both recurrence-derived osteoblast candidates and an outcome-associated molecular subtype. The CGREF1–COL13A1–KIF25–RPL35AP7 score achieved an apparent C-index of 0.814 and 1-, 3-, and 5-year areas under the receiver operating characteristic curve of 0.855, 0.852, and 0.845. RiskScore remained associated with overall survival after adjustment for age, sex, and metastatic status (hazard ratio 2.413, 95% confidence interval 1.769–3.290). Higher scores tracked lower B-lineage, neutrophil, and fibroblast signals. Gene-directed treatments reduced wound closure in both cell lines. COL13A1-directed treatment reduced colony formation in both cell lines, and KIF25-directed treatment reduced colony formation in HOS cells (Dunnett-adjusted P < 0.05). Conclusion This four-gene signature connects osteosarcoma outcome stratification with an immunometabolic and stromal context and identifies COL13A1 and KIF25 as candidates for focused functional investigation.