Identification of Prognostic RBPs in Osteosarcoma

Abstract

Osteosarcoma often occurs in children and adolescents and causes poor prognosis. The role of RNA-binding proteins (RBPs) in malignant tumors has been elucidated in recent years. Our study aims to identify key RBPs in osteosarcoma that could be prognostic factors and treatment targets. GSE33382 dataset was downloaded from Gene Expression Omnibus (GEO) database. RBPs extraction and differential expression analysis was performed. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis were performed to explore the biological function of differential expression RBPs. Moreover, we constructed Protein-protein interaction (PPI) network and obtained key modules. Key RBPs were identified by univariate Cox regression analysis and multiple stepwise Cox regression analysis combined with the clinical information from Therapeutically Applicable Research to Generate Effective Treatments (TARGET) database. Risk score model was generated and validated by GSE16091 dataset. A total of 38 differential expression RBPs was identified. Go and KEGG results indicated these RBPs were significantly involved in ribosome biogenesis and mRNA surveillance pathway. COX regression analysis showed DDX24, DDX21, WARS and IGF2BP2 could be prognostic factors in osteosarcoma. Spearman's correlation analysis suggested that WARS might be important in osteosarcoma immune infiltration. In conclusion, DDX24, DDX21, WARS and IGF2BP2 might play key role in osteosarcoma, which could be therapuetic targets for osteosarcoma treatment.

Keywords: RBPs; bioinformation; immune infiltration; osteosarcoma; prognosis; risk score model.

Figure 1.

Identification of differentially expressed RBPs from dataset. (A) Volcano plot of differentially expressed RBPs analysis for GSE33382. The red dots represent for up-regulated RBPs, the green dots represent for down-regulated RBPs. The thresholds for identification are shown as blue dotted lines [|log2 fold-change (FC)| > 1 and adj. P < 0.05] . (B) Heatmap of differentially expressed RBPs.

Figure 2.

GO enrichment analysis for differentially expressed RBPs.

Figure 3.

PPI network and key modules. (A) The whole PPI network. (B) Subnetwork of module 1. (C) Subnetwork of module 2. (D) Subnetwork of module 3.

Figure 4.

Cox regression analysis of key RBPs from PPI network. (A) Univariate Cox regression analysis. (B) Stepwise multiple Cox regression analysis.

Figure 5.

Survival curve and ROC curve analyses of TARGET and GEO cohort. (A) Survival curve of risk score in TARGET cohort. (B) ROC curve of risk score in TARGET cohort. (C) Survival curve of risk score in GEO cohort. (D) ROC curve of risk score in GEO cohort.

Figure 6.

Expression heatmap, risk score distribution and survival status in TARGET cohort and GEO cohort. (A) Expression heatmap of hub RBPs in TARGET cohort. (B) Risk score distribution in TARGET cohort. (C) Survival status in TARGET cohort. (D) Expression heatmap of hub RBPs in GEO cohort. (E) Risk score distribution in GEO cohort. (F) Survival status in GEO cohort.

Figure 7.

Nomogram for predicting 1-, 3-, and 5-year OS of osteosarcoma patients in TARGET cohort.

Figure 8.

Spearman’s correlation analysis between hub RBPs and tumor immune infiltration. (A) Correlation between DDX21, WARS and tumor purity. (B) Correlation between DDX24, WARS and immune cells.

Figure 9.

Expression of the 4 hub RBPs in osteosarcoma cell lines (MG-63 and MNNG/HOS) and osteoblast cell line (hFOB 1.19). *P < 0.05.