Machine-Learning-Based m5C Score for the Prognosis Diagnosis of Osteosarcoma

Abstract

Background. Osteosarcoma is a common and highly metastatic malignant tumor, and m5C RNA methylation regulates various biological processes. The purpose of this study was to explore the prognostic role of m5C in osteosarcoma using machine learning. Methods. Osteosarcoma gene data and the corresponding clinical information were downloaded from the GEO database. Machine learning methods were used to screen m5C-related genes and construct m5C scores. In addition, the clusterProfiler package was used to predict the m5C-related functional pathways. xCell and CIBERSORT were used to calculate the immune microenvironment cells. GSVA was applied to analyze different categories of m5C genes, and the correlation between the GSVA and m5C scores was evaluated. Results. Twenty m5C genes were identified, and 54 related genes were screened. The m5C score was constructed based on the PCA score. With an increase in the m5C score, the expression of m5C genes and their related genes changed. Functional analysis indicated that the focal adhesion, cell-substrate adherens junction, cell adhesion molecule binding, and E2F targets might change with the m5C score. The naive B cells and CD4+ memory T cell also changed with the m5C score. The results of the correlation analysis showed that the m5C score was significantly correlated with the reader and eraser genes. Conclusion. The m5C score might be a prognostic index for osteosarcoma.