Table 2. Metrics for Different Machine Learning Algorithms Using the Judred and Mordred Data sets Comparatively^a.

	MSE	MAE	R2	MCC	AUC
Judred
SVMRBF	0.0152	0.0941	0.93	0.88	0.99
linear SVM	0.0529	0.1884	0.75	0.69	0.96
gradient boosting regressor	0.0146	0.0890	0.93	0.87	0.99
elastic net	0.0533	0.1875	0.75	0.70	0.96
random forest	0.0478	0.1699	0.78	0.79	0.97
ridge	0.0527	0.1864	0.75	0.70	0.96
multi-layer perceptron	0.0144	0.0892	0.93	0.86	0.99
stochastic gradient descent	0.0513	0.1835	0.76	0.70	0.96
decision tree	0.0150	0.0899	0.93	0.86	0.99
Mordred
SVMRBF	0.0072	0.0673	0.97	0.92	0.99
linear SVM	0.0232	0.1225	0.89	0.82	0.98
gradient boosting regressor	0.0088	0.0708	0.96	0.88	0.99
elastic net	0.0268	0.1310	0.87	0.78	0.98
random forest	0.0475	0.1693	0.78	0.73	0.97
ridge	0.0242	0.1244	0.87	0.82	0.98
multi-layer perceptron	0.0065	0.0609	0.97	0.90	0.99
stochastic gradient descent	0.0248	0.1270	0.88	0.82	0.99
decision tree	0.0177	0.0933	0.92	0.85	0.99

^aAn 80–20 train-test split was used to determine the best models according to five metrics; for MCC, a cut-off of AP = 2.0 was used. Receiver operating characteristic curves as well as testing on a 66–34 train-test split are included in the Supporting Information, Figure S3 and Table S3.