Fig. 2From: A systematic approach to identify therapeutic effects of natural products based on human metabolite informationAUROC value of models generated from different feature sets. SVM models are trained with single features and compared with the model trained with the whole feature sets. a For random test set, SVM model trained only with structure feature shows as high AUROC as the model trained with the whole feature sets. b For structurally similar test set (Tanimoto score ≥ 0.77), all features contribute to improve the overall performanceBack to article page