Construction and Validation of a Macrophage-Associated Risk Model for Predicting the Prognosis of Osteosarcoma

Kang-Wen Xiao; Zhi-Bo Liu; Zi-Hang Zeng; Fei-Fei Yan; Ling-Fei Xiao; Jia-Li Li; Lin Cai

doi:10.1155/2021/9967954

Construction and Validation of a Macrophage-Associated Risk Model for Predicting the Prognosis of Osteosarcoma

J Oncol. 2021 Jun 2:2021:9967954. doi: 10.1155/2021/9967954. eCollection 2021.

Authors

Kang-Wen Xiao¹, Zhi-Bo Liu¹, Zi-Hang Zeng², Fei-Fei Yan¹, Ling-Fei Xiao¹, Jia-Li Li², Lin Cai¹

Affiliations

¹ Department of Orthopedics, Zhongnan Hospital of Wuhan University, Wuhan, Hubei430071, China.
² Department of Radiation and Medical Oncology, Zhongnan Hospital of Wuhan University, Wuhan, Hubei430071, China.

Abstract

Background: Osteosarcoma is one of the most common bone tumors among children. Tumor-associated macrophages have been found to interact with tumor cells, secreting a variety of cytokines about tumor growth, metastasis, and prognosis. This study aimed to identify macrophage-associated genes (MAGs) signatures to predict the prognosis of osteosarcoma.

Methods: Totally 384 MAGs were collected from GSEA software C7: immunologic signature gene sets. Differential gene expression (DGE) analysis was performed between normal bone samples and osteosarcoma samples in GSE99671. Kaplan-Meier survival analysis was performed to identify prognostic MAGs in TARGET-OS. Decision curve analysis (DCA), nomogram, receiver operating characteristic (ROC), and survival curve analysis were further used to assess our risk model. All genes from TARGET-OS were used for gene set enrichment analysis (GSEA). Immune infiltration of osteosarcoma sample was calculated using CIBERSORT and ESTIMATE packages. The independent test data set GSE21257 from gene expression omnibus (GEO) was used to validate our risk model.

Results: 5 MAGs (MAP3K5, PML, WDR1, BAMBI, and GNPDA2) were screened based on protein-protein interaction (PPI), DGE, and survival analysis. A novel macrophage-associated risk model was constructed to predict a risk score based on multivariate Cox regression analysis. The high-risk group showed a worse prognosis of osteosarcoma (p < 0.001) while the low-risk group had higher immune and stromal scores. The risk score was identified as an independent prognostic factor for osteosarcoma. MAGs model for diagnosis of osteosarcoma had a better net clinical benefit based on DCA. The nomogram and ROC curve also effectively predicted the prognosis of osteosarcoma. Besides, the validation result was consistent with the result of TARGET-OS.

Conclusions: A novel macrophage-associated risk score to differentiate low- and high-risk groups of osteosarcoma was constructed based on integrative bioinformatics analysis. Macrophages might affect the prognosis of osteosarcoma through macrophage differentiation pathways and bring novel sights for the progression and prognosis of osteosarcoma.