. 2022 Oct 4;62(20):4852–4862. doi: 10.1021/acs.jcim.2c00715

Table 6. Comparison of the Performance of MFBERT with Other State-of-the-Art Models from the Literature^a.

	MFBERT (+FFNN)	MFBERT (best)	ChemBERTa	SSLP-FPs (best)	MoleculeNet (best)	MolBERT (fine-tune)
BBBP*	0.762	0.762	0.663	0.741	0.729	0.762
ClinTox*	0.912	0.912	0.842	0.963 ± 0.044	0.907	0.912
HIV*	0.765	0.765	0.643	0.601	0.792	0.783
ESOL^†	1.153 ± 0.141	0.423 ± 0.501		0.915 ± 0.01	0.58	0.531 ± 0.04
FreeSolv^†	1.180 ± 0.440	0.624 ± 0.130		0.891 ± 0.075	1.15	0.948 ± 0.33
lipophilicity^†	0.868 ± 0.021	0.366 ± 0.032		0.717 ± 0.029	0.655	0.561 ± 0.03

*AUCROC score; ^†root mean squared error (RMSE) score.

Table 6. Comparison of the Performance of MFBERT with Other State-of-the-Art Models from the Literaturea.