Construction and evaluation of a prognostic model based on the expression of the metabolism-related signatures in patients with osteosarcoma.

Abstract

Background: The aim of this study was to screen three major substance metabolism-related genes and establish a prognostic model for osteosarcoma.

Methods: RNA-seq expression data for osteosarcoma were downloaded from The Cancer Genome Atlas (TCGA) and GEO databases. Differentially expressed (DE) RNAs were selected, followed by the selection of metabolic-related DE mRNAs. Using Cox regression analysis, prognostic DE RNAs were identified to construct a prognostic model. Subsequently, independent prognostic clinical factors were screened, and the functions of the long non-coding RNAs (lncRNAs) were analyzed. Finally, the expression of signature genes was further tested in osteosarcoma cells using quantitative reverse transcription quantitative real-time polymerase chain reaction (qRT-PCR) and western blotting.

Results: A total of 432 DE RNAs, comprising 79 DE lncRNAs and 353 DE mRNAs were obtained, and then 107 metabolic-related DE mRNAs. Afterwards signature genes (LINC00545, LINC01537, FOXC2-AS1, CYP27B1, PFKFB4, PHKG1, PHYKPL, PXMP2, and XYLB) served as optimal combinations, and a prognostic score model was successfully proposed. Three verification datasets (GSE16091, GSE21257, and GSE39055) showed that the model had high specificity and sensitivity. In addition, two independent prognostic clinical factors (age and tumor metastasis) were identified. Finally, the concordance rate between the in silico analysis, qRT-PCR, and western blotting analysis was 88.89% (8/9), suggesting the robustness of our analysis.

Conclusions: The prognostic model based on the nine signature genes accurately predicted the prognosis of patients with osteosarcoma; CYP27B1, PFKFB4, PHKG1, PHYKPL, PXMP2, and XYLB may serve as metabolism-related biomarkers in osteosarcoma.