Abstract
BACKGROUND/AIMS: Osteosarcoma (OS), a malignant tumor originating in the bone or cartilage, primarily affects children and adolescents. Notably, glycolysis is the main target for metabolic programming to ensuring the energy supply for cancer. This study aimed to establish a glycolysis-related gene (GRG) risk signature in OS to comprehensively assessing the pathogenic, prognosis, and their application in predicting drug response. METHODS: mRNA expression profiles were acquired from the Gene Expression Omnibus (GEO, GSE16091, GSE39058, and GSE21257). Using the non-negative matrix factorization (NMF) algorithm, patients with OS were stratified into distinct subgroups based on 288 GRGs identified through univariable Cox analysis. Univariate Cox regression analysis of differentially expressed genes (DEGs) between the molecular clusters was conducted to establish a risk signature comprising GRGs in OS. The prognostic efficacy of this risk signature was assessed via Kaplan-Meier curve analysis and Cox regression, evaluating its independence as a prognostic indicator. Additionally, the predictive potential of the risk model for drug response was evaluated using the "OncoPredict" package. Furthermore, the distribution of immune cell types in single-cell RNA sequencing (scRNA-seq) data was examined in correlation with the four identified GRGs risk signatures, followed by validation of expression levels in vitro using RT-PCR. RESULTS: Patients diagnosed with OS were categorized into two distinct molecular subgroups, exhibiting notable variations in prognosis and tumor microenvironment. Univaria te Cox regression analysis was employed to identify four GRGs, namely chondroitin sulfate glucuronyltransferase (CHPF), Ras-related GTP-binding protein D (RRAGD), nucleoprotein TPR (TPR), and versican core protein (VCAN), which constitute a prognostic signature for patients with OS. This signature demonstrated robust prognostic value, as corroborated by Kaplan-Meier, univariate, and multivariate Cox regression analyses. Significant differences in tumor microenvironment immune infiltration (such as B cells, monocytes) were observed between molecular subgroups. Moreover, a significant disparity in drug sensitivity to AZD8055, paclitaxel, and PD0325901 was noted between the high-risk and low-risk cohorts, and the established four-gene risk signature served as dependable prognostic indicators in the validation cohort, confirmed at the cellular level through external dataset validation and reverse transcription quantitative PCR (RT-qPCR) experiments. CONCLUSION: A risk signature based on GRGs was established for OS, exhibiting robust predictive efficacy for prognostic assessment, and offering significant clinical utility for the prognosis of OS.