Abstract Background Glioma is the most commonly diagnosed malignant and aggressive brain cancer in adults.Traditional researches mainly explored the expression profile of glioma at cell-population level, but ignored the heterogeneity and interactions of among glioma cells.Methods Here, we firstly analyzed the single-cell RNA-seq (scRNA-seq) data of 6341 glioma cells using manifold learning and identified neoplastic and healthy cells infiltrating in tumor microenvironment.We systematically revealed cell-to-cell interactions inside gliomas based on corresponding scRNA-seq and TCGA RNA-seq data.
Results A Range total of 16 significantly correlated autocrine ligand-receptor signal pairs inside neoplastic cells were identified based on the scRNA-seq and TCGA data of glioma.Furthermore, we explored the intercellular communications between cancer stem-like cells (CSCs) and macrophages, and identified 66 ligand-receptor pairs, some Resistance Bands of which could significantly affect prognostic outcomes.An efficient machine learning model was constructed to accurately predict the prognosis of glioma patients based on the ligand-receptor interactions.Conclusion Collectively, our study not only reveals functionally important cell-to-cell interactions inside glioma, but also detects potentially prognostic markers for predicting the survival of glioma patients.