Construction of a Survival Prediction Model and Molecular Subtype Analysis Based on TARGET Osteosarcoma Multi-omics Data

Background Osteosarcoma is the most common malignant bone tumor in adolescents and young adults, characterized by high heterogeneity and poor prognosis. Despite advances in multimodal therapy, reliable prognostic biomarkers and molecular classifications remain limited. Methods Multi-omics data (RNA-seq, miRNA expression, CNV, DNA methylation, and clinical variables) were obtained from the TARGET osteosarcoma cohort. After normalization and batch effect correction, prognostic features were identified using univariate Cox regression and LASSO-Cox modeling. Prognostic risk scores were constructed and validated with time-dependent ROC, C-index, Brier score, and calibration analyses. Consensus clustering and multi-omics factor analysis (MOFA) were applied to define molecular subtypes. Results Transcriptomic features exhibited the strongest prognostic performance (C-index = 0.726), outperforming clinical-only or integrated multi-omics models. A minimal LASSO-Cox gene signature stratified patients into high- and low-risk groups with significantly different survival outcomes (log-rank p  < 0.001). Consensus clustering identified two robust molecular subtypes, which displayed distinct survival patterns, clinical characteristics, and multi-omics profiles. Mechanistic analysis suggested that dysregulation of PI3K/AKT, Wnt/β-catenin, and TGF-β signaling, combined with an immunosuppressive tumor microenvironment, underlies poor prognosis in the high-risk subtype. Conclusions We developed a robust prognostic model and identified biologically distinct molecular subtypes in osteosarcoma using integrative multi-omics analysis. These findings provide new insights into tumor heterogeneity and potential precision therapeutic strategies.