Fig. 4

The 32-gene-based model achieves a better performance than other models. A SHAP value (feature importance score) of the 32 genes. B The histogram comparison of the Accuracy, AUC, Recall, Precision, F1 and Kappa of two XGBoost-based classification models, those values range between 0 and 1 (0, poor performance; 1, good performance). C-H Confusion matrix of 32, 54, 30, 21, Path 1-2-3, and Top SHAP value gene-based XGBoost classification models with samples that were not used for training and validation. I Confusion matrix of the 32-gene-based Random Forest classification with samples that were not used for training and validation. Confusion matrix detailing the true positive (right lower), true negative (left upper), false positive (right upper), and false negative (left lower) predictions from XGBoost-based classification model. Accuracy = (true positive + true negative) / total