Figure 7.
Results of various classic machine learning algorithms and deep learning on the 153 compound TB test set using different evaluation metrics. The green bar represents the highest ROC AUC for each category. Machine learning algorithms compared include: Logistic Regression = LR, Adaboost Decision Trees = ADA, Random Forest = RF, Bernoulli Naïve-Bayesian = BNB and Support Vector Machines = SVM, DNN = Deep Neural Network with variable layers. Further details can be found in Supplemental Tables 2–6.