. Author manuscript; available in PMC: 2023 Jan 31.

Published in final edited form as: Chem Res Toxicol. 2020 Dec 21;34(2):495–506. doi: 10.1021/acs.chemrestox.0c00322

Table 5:

The performance of DILI models using cross-validation with random splitting (higher is better). Best score is marked as bold.

	Circular-fp	Neural-fp	MPNN	DMPNN	MV-GNN^cross	Property Augmentation with Tox-DILI
Accuracy	0.688_±0.051	0.704_±0.091	0.738_±0.094	0.750_±0.098	0.788_±0.077	0.814 _±0.047
Sensitivity	0.364_±0.125	0.647_±0.091	0.727_±0.133	0.728_±0.135	0.762_±0.105	0.768 _±0.100
Specificity	0.879 _±0.086	0.740_±0.087	0.752_±0.129	0.764_±0.172	0.809_±0.092	0.849_±0.097
F1-score	0.485_±0.091	0.615_±0.106	0.666_±0.124	0.681_±0.095	0.721_±0.105	0.753 _±0.063
MCC	0.289_±0.130	0.381_±0.191	0.473_±0.202	0.499_±0.179	0.562_±0.178	0.621 _±0.114
ROC-AUC	0.738_±0.056	0.753_±0.093	0.833_±0.075	0.832_±0.068	0.866_±0.055	0.882 _±0.031