Search

Esophageal squamous cell carcinoma (ESCC) is a malignant tumor of the digestive system in the esophageal squamous epithelium. Many studies have linked esophageal cancer (EC) to the imbalance of oral microecology. In this work, different machine learning (ML) models including Random Forest (RF), Gaussian mixture model (GMM), K-nearest neighbor (KNN), logistic regression (LR), support vector machine (SVM) and extreme gradient boosting (XGBoost) based on Genetic Algorithm (GA) optimization was developed to predict the relationship between salivary flora and ESCC by combining the relative abundance data of Bacteroides, Firmicutes, Proteobacteria, Fusobacteria and Actinobacteria in the saliva of patients with ESCC and healthy control. The results showed that the XGBoost model without parameter optimization performed best on the entire dataset for ESCC diagnosis by cross-validation (Accuracy = 73.50%). Accuracy and the other evaluation indicators, including Precision, Recall, F1-score and the area under curve (AUC) of the receiver operating characteristic (ROC), revealed XGBoost optimized by the GA (GA-XGBoost) achieved the best outcome on the testing set (Accuracy = 89.88%, Precision = 89.43%, Recall = 90.75%, F1-score = 90.09%, AUC = 0.97). The predictive ability of GA-XGBoost was validated in phylum-level salivary microbiota data from ESCC patients and controls in an external cohort. The results obtained in this validation (Accuracy = 70.60%, Precision = 46.00%, Recall = 90.55%, F1-score = 61.01%) illustrate the reliability of the predictive performance of the model. The feature importance rankings obtained by XGBoost indicate that Bacteroides and Actinobacteria are the two most important factors in predicting ESCC. Based on these results, GA-XGBoost can predict and diagnose ESCC according to the relative abundance of salivary flora, providing an effective tool for the non-invasive prediction of esophageal malignancies.

Esophageal squamous cell carcinoma (ESCC) is a malignant tumor of the digestive system in the esophageal squamous epithelium. Many studies have linked esophageal cancer (EC) to the imbalance of oral microecology. In this work, different machine learning (ML) models including Random Forest (RF), Gaussian mixture model (GMM), K-nearest neighbor (KNN), logistic regression (LR), support vector machine (SVM) and extreme gradient boosting (XGBoost) based on Genetic Algorithm (GA) optimization was developed to predict the relationship between salivary flora and ESCC by combining the relative abundance data of Bacteroides , Firmicutes , Proteobacteria , Fusobacteria and Actinobacteria in the saliva of patients with ESCC and healthy control. The results showed that the XGBoost model without parameter optimization performed best on the entire dataset for ESCC diagnosis by cross-validation (Accuracy = 73.50%). Accuracy and the other evaluation indicators, including Precision, Recall, F1-score and the area under curve (AUC) of the receiver operating characteristic (ROC), revealed XGBoost optimized by the GA (GA-XGBoost) achieved the best outcome on the testing set (Accuracy = 89.88%, Precision = 89.43%, Recall = 90.75%, F1-score = 90.09%, AUC = 0.97). The predictive ability of GA-XGBoost was validated in phylum-level salivary microbiota data from ESCC patients and controls in an external cohort. The results obtained in this validation (Accuracy = 70.60%, Precision = 46.00%, Recall = 90.55%, F1-score = 61.01%) illustrate the reliability of the predictive performance of the model. The feature importance rankings obtained by XGBoost indicate that Bacteroides and Actinobacteria are the two most important factors in predicting ESCC. Based on these results, GA-XGBoost can predict and diagnose ESCC according to the relative abundance of salivary flora, providing an effective tool for the non-invasive prediction of esophageal malignancies.