Abstract
BACKGROUND: Currently, there is no research on building osteosarcoma (OS) prognostic models based on single-cell RNA sequencing (scRNA-seq) and anoikis-related genes (ARGs). METHODS: Differential genes between osteoblasts cells and osteosarcoma cells were identified using scRNA-seq, and ARGs were determined by Genecard database. Lasso regression was employed to investigate hub genes and construct the model based on TARGET. Kaplan-Meier survival analysis was applied to compare the survival differences. ROC curves were used to evaluate the predictive performance of the model. CIBERSORT and ESTIMATE algorithms were conducted to calculate immune cell infiltration abundance. Finally, qRT-PCR and immunohistochemistry experiments were conducted to validate the results. RESULTS: A predictive model containing four modeling genes (MYC, BNIP3, IGFBP5, and SPP1) was successfully constructed, with AUC values of 0.836, 0.837, and 0.836 for 1-, 3-, and 5-year patient prognosis, respectively. Importantly, the model also showed good predictive value in two validation set. The infiltration of immune cells in different risk groups showed significant differences. The modeling genes were associated with the expression of various immune checkpoints and the response to immune therapy. qRT-PCR showed MYC, BNIP3, IGFBP5, and SPP1 substantially exhibited a trend of high expression in osteosarcoma cells. Immunohistochemistry suggested that in osteosarcoma patients with poorer prognosis, the expression of these four hub genes was significantly elevated. CONCLUSIONS: We have developed an effective model for predicting osteosarcoma prognosis and immune response, which may provide valuable insights for osteosarcoma prognostic evaluation and immune therapy strategies.