Immune Cell Infiltration Score and Predicting Model among Lower-grade Gliomas Based on Immunogenomic Clusters

Background and objectives : Recurrent malignancies had become a significant problem for the treatment of gliomas. Though immunotherapy was regarded as a possible solution, verification of immune-checkpoint inhibitors in multiple clinical cases failed. We aimed to explore target genes for immunotherapy and evaluate the genes through new scoring criteria. Methods : We firstly classified the patients through k-means clustering in immune cell level and gene level. Differential prognostic genes were weighted through principal component analysis (PCA) and Boruta algorithm. The comprehensive scoring of each component was defined as the ICI score. We further analyzed the relationship between ICI score and various clinical factors and prognosis. Moreover, a nomogram based on the ICI scores was built and validated through both internal and external validation. Results : The study cohort finally enrolled 495 patients. We identified a list of differential genes of which the expression level was closely related to prognosis. ICI scores were calculated for each case according to the expression level of prognostic genes. The cases in high ICI score group showed significantly lower survival probabilities (log-rank test; p<0.001). We further built a nomogram model based on the ICI scores. The area under the receiver operating characteristic (ROC) curve (AUC) values of the nomogram were 0.851, 0.86, and 0.768 in internal validation, and 0.725, 0.744, and 0.735 for 1-year, 3-year and 5-year truncation time in external validation. Conclusions : The differential genes were listed for further studies. ICI score had been confirmed to be closely related to prognosis and even genome instability. And the nomogram model based on ICI scores showed the feasibility in clinical practice in both internal and external validation.