Additional file 1: Table S1 Model performance on threefold cross validations using M-data. Table S2 Model performance on S-data trained by M-data with 5 times runs. Table S3 Substrates in S-data. Table S4 Model performance on substrates in S-data. Table S5 Model performance of all combinations of uncertainty estimation methods in GROVER-BBBp model. Table S6 A list of antitumor SMIs and corresponding prediction results in GROVER-BBBp model. Table S7 Prediction results for external dataset collected from Colclough and Kim’s works in GROVER-BBBp model. Table S8 The details of 9 drug-like datasets from admetSAR. Table S9 Model performance on fivefold cross-validations of 9 drug-like datasets in GROVER model. Table S10 Model performance on test set of 9 drug-like datasets in GROVER model. Fig. S1 Prediction performance by introducing different uncertainty estimation methods in 9 GROVER models. Table S11 Model performance of various combination of uncertainty estimation methods on fathead minnow toxicity dataset in GROVER model. Table S12 Model performance of various combination of uncertainty estimation methods on tetrahymena pyriformis toxicity dataset in GROVER model. Table S13 Model performance of various combination of uncertainty estimation methods on AMES mutagenicity dataset in GROVER model. Table S14 Model performance of various combination of uncertainty estimation methods on hERG inhibitor (II) dataset in GROVER model. Table S15 Model performance of various combination of uncertainty estimation methods on CYP3A4 substrates dataset in GROVER model. Table S16 Model performance of various combination of uncertainty estimation methods on CYP1A2 inhibitor dataset in GROVER model. Table S17 Model performance of various combination of uncertainty estimation methods on Caco-2 permeability dataset in GROVER model. Table S18 Model performance of various combination of uncertainty estimation methods on P-gp inhibitor (I) dataset in GROVER model. Table S19 Model performance of various combination of uncertainty estimation methods on P-gp inhibitor (II) dataset in GROVER model.
