Combined bioinformatics and machine learning methodologies reveal prognosis-related ceRNA network and propose ABCA8, CAT, and CXCL12 as independent protective factors against osteosarcoma.

Background: Aberrant circular RNA (circRNA) acts as an oncogene or suppressor during neoplasm initiation and development. However, the functions of most circRNAs in osteosarcoma (OS) remain unclear.
Objectives: We aimed to investigate the expression, molecular functions and mechanisms underlying circRNAs in OS.
Material and Methods: Network interaction, pathway enrichment and regression analyses were performed to determine differentially expressed (DE) circRNAs, microRNAs (miRNAs) and messenger RNAs (mRNAs). We constructed competitive endogenous RcodeNA (ceRNA) networks and integrated patient clinical data to analyze the relationship between the networks and prognosis. The circRNA, miRNA and mRNA data were retrieved from Gene Expression Omnibus (GEO) microarray datasets. A circRNA-miRNA-mRNA interaction network was established and visualized using miRNet. Protein interactions were investigated using STRING and Cytoscape, and hub genes were identified using the MCODE plug-in. Gene Ontology, Kyoto Encyclopedia of Genes and Genomes (KEGG) and Reactome pathway analyses were performed to determine the DEmRNAs. LIMMA and RobustRankAggreg were used to screen for DERNAs. Node genes in the interaction network were analyzed using least absolute shrinkage and selection operator (LASSO) and Cox regression to obtain OS-related ceRNA networks.
Results: We identified 9 DEcircRNAs, 243 DEmiRNAs and 211 DEmRNAs. We found that a ceRNA subnetwork, based on 1 circRNA, 1 miRNA and 8 mRNAs, was closely associated with OS prognosis. Integrating the proportional hazards model and survival analysis revealed 3 independent protective factors: adenosine triphosphate (ATP)-binding cassette sub-family A member 8 (ABCA8), catalase (CAT) and C-X-C motif chemokine ligand 12 (CXCL12).
Conclusions: Our study provides novel insights into circRNA-related ceRNA networks and identifies potential prognostic biomarkers of OS.