Objective To establish a machine learning model using radiomics features extracted from preoperative MRI to predict the expression levels of the key immune checkpoint， programmed death receptor-1 （PD-1） gene， in patients with glioma. Methods From January to August 2024， 133 patients were retrospectively collected from The Cancer Genome Atlas（ TCGA） database. Using the random sampling "sample function" in R software， patients were randomly assigned in a 2∶1 ratio to training group（ 89 cases） and validation group（ 44 cases）. Feature selection and predictive model construction were performed using the LASSO algorithm. The predictive performance of the model was evaluated on the training and validation sets using the receiver operating characteristic（ ROC） curve. The clinical practicality of the model was assessed through decision curve analysis. Results Six radiomics features were selected using the LASSO algorithm， and a predictive model was constructed. Six radiomics features were selected using the LASSO algorithm， and a predictive model was constructed. ROC curve analysis on the training and validation sets yielded C-index values of 0.815 and 0.728， respectively， with accuracy rates of 80.9% and 72.7%. Conclusions The expression level of PD-1， a key immune checkpoint in glioma patients， can be predicted using preoperative MRI data and radiomics technology.