Machine learning-driven SLC prognostic signature for glioma: predicting survival and immunotherapy response

Yu'e Liu^*

• Boston Children's Hospital, Harvard Medical School, Boston, United States

Gliomas, the most common and aggressive primary brain tumors, exhibit significant heterogeneity and poor prognosis. Despite advancements in treatment, patient outcomes remain unsatisfactory due to therapeutic resistance and tumor recurrence. Identifying novel molecular biomarkers is crucial for improving prognosis and developing more effective therapies. In this study, we constructed a solute carrier family prognostic signature (SLCFPS) for gliomas by applying univariate Cox regression and machine learning algorithms to five independent glioma cohorts. The prognostic performance of SLCFPS was assessed using Kaplan-Meier survival analysis, concordance index (C-index), and receiver operating characteristic (ROC) curve analysis. The immune landscape, immunotherapy response, and drug sensitivity were further explored through bioinformatics tools such as ESTIMATE, xCell, TIDE, and drug response correlation analysis. SLCFPS effectively stratified glioma patients into high-and low-risk groups, with higher scores correlating with poorer survival outcomes. The model demonstrated superior predictive performance compared to existing glioma prognostic models and was associated with an immunosuppressive tumor microenvironment, as well as upregulated immune checkpoints, suggesting implications for immunotherapy response. Furthermore, SLCFPS correlated with drug sensitivity, highlighting potential therapeutic options. Overall, SLCFPS is a robust biomarker for glioma prognosis and treatment response, offering insights into tumor immunity and facilitating personalized therapeutic strategies.