Abstract
Traditional glioma diagnostic methods have limitations, while liquid biopsy is a promising non-invasive option. This study developed the glioma-related cell signature (GRCS), a prediction model that integrates machine learning with biological insights. Trained on tumor-educated platelet samples, the GRCS model demonstrated consistent performance across validation cohorts comprising platelet, extracellular vesicle, and tumor tissue specimens. The GRCS score showed significant associations with patient age, histological grade, survival outcome, and mutational landscape. Moreover, the GRCS model effectively distinguished responses to bevacizumab and immunotherapy and identified potential candidates for combination therapies. Furthermore, a miRNA-based simplified GRCS model (GRCSS) was developed and validated across different specimen cohorts, demonstrating its robust diagnostic and prognostic capabilities in glioma. This work highlights the potential of GRCS as a versatile tool for personalized glioma management across multiple biopsy specimen types.