A computationally derived five-gene risk signature from broad mitophagy- and tumor metabolism-associated candidate screening in osteosarcoma

Osteosarcoma (OS) is a primary bone cancer affecting adolescents and young adults, with high mortality and recurrence rates. This study integrated publicly available transcriptomic, clinical, and somatic mutation data to derive an exploratory prognostic risk signature. A broad pool of 3,745 analyzable candidate genes was obtained by matching genes collected through mitophagy- and tumor metabolism-associated searches of GeneCards and relevant literature to an OS expression matrix. Exploratory differential expression analysis identified 25 candidate genes, including 15 upregulated and 10 downregulated genes. Weighted gene co-expression network analysis (WGCNA) narrowed these to 12 module genes. Survival modeling in the TARGET-OS discovery cohort, including least absolute shrinkage and selection operator (LASSO) regression, subsequently identified a five-gene signature comprising BBS4, CTNND2, EPHA3, MAPK10, and MYC. These genes represent statistically prioritized candidate prognostic biomarkers; their direct functional contributions to mitophagy or tumor metabolism in OS were not established. The signature stratified overall survival in the TARGET-OS discovery cohort ( p < 0.001), but significant survival stratification was not reproduced in GSE16091, GSE21257, or GSE39055 ( p = 0.35, 0.19, and 0.61, respectively). The model genes showed risk-group-specific correlations with computationally estimated immune-cell features. Single-cell analysis described their cell-type-specific expression, while drug-sensitivity analysis identified differences in computationally predicted drug responses. These findings provide hypotheses for further investigation rather than evidence of direct immune regulation or clinical treatment selection. The five-gene signature therefore remains an exploratory finding requiring independent prognostic and functional validation.