. 2023 Feb 3;13(7):4623–4640. doi: 10.1039/d2ra07007c

Influence of incorporating intermediate-activity ligands on the success criteria of the resulting ML models.

Learner	Intermediate activity class	Accuracy		Cohen's Kappa
Learner	Intermediate activity class	L20% out^a	Testing^b	L20% out^a	Testing^b
XGboost	With	0.60	0.60	0.35	0.35
XGboost	Without	0.77	0.76	0.49	0.46
RF	With	0.60	0.58	0.34	0.31
RF	Without	0.77	0.76	0.47	0.45

L20% out: leave 20% out cross-validation for accuracy and Cohen's Kappa.

Testing: accuracy and Cohen's Kappa determined against the testing set (marked with a in Table S1).