. Author manuscript; available in PMC: 2020 Feb 25.

Published in final edited form as: J Chem Inf Model. 2019 Feb 12;59(2):689–701. doi: 10.1021/acs.jcim.9b00020

Table 3.

Virtual Screening Results on 10 DUD Targets^a

			Method
Target	Average Similarity to Cognate Ligand		BCL::MolAlign			Maximum Common Substructure Similarity			PL-PatchSurfer2
	Actives	Decoys	AUC	ROCE_1%	ROCE_5%	AUC	ROCE_1%	roce_5%	AUC	ROCE_1%	roce_5%
ACHE	0.45 ±0.21	0.20 ±0.07	0.9	71.6	15.3	0.9	95.6	28.8	0.7	28.6	7.2
CDK2	0.19 ±0.04	0.20 ± 0.04	0.5	2.1	3.0	0.4	0.0	0.4	0.5	109.9	6.6
ERA	0.28 ±0.12	0.18 ±0.05	0.9	49.0	20.2	0.8	172.2	12.2	0.5	12.4	4.7
HIVPR	0.13 ±0.04	0.12 ±0.03	0.7	1.9	3.2	0.6	2.0	0.4	0.7	121.8	11.9
NA	0.34 ±0.14	0.14 ±0.07	0.9	107.0	19.9	0.9	10.3	25.4	0.5	29.6	3.1
P38	0.16 ±0.09	0.15 ±0.04	0.7	14.5	7.4	0.3	14.5	2.1	0.6	18.6	5.3
PPARG	0.59 ±0.29	0.15 ±0.05	0.9	62.7	29.1	1.0	656.0	41.3	0.5	0.0	0.5
RXRA	0.69 ±0.13	0.18 ±0.07	1.0	1.0	48.0	1.0	145.0	52.7	0.7	14.5	5.8
SRC	0.14 ±0.03	0.18 ±0.06	0.6	1.7	2.1	0.2	0.0	0.0	0.6	148.3	10.0
THROMBIN	0.15 ±0.06	0.14 ±0.05	0.8	1.6	1.6	0.7	1.5	1.3	0.6	10.3	6.2
Median	--	--	0.8	8.3	11.4	0.8	12.4	7.2	0.6	23.6	6.0

Evaluation of the virtual screening performance of BCL::MolAlign compared to maximum common substructure similarity and PL-PatchSurfer2 on 10 targets originally reported by Shin et al.⁴⁴ Similarity was measured as the maximum common connected substructure Tanimoto score. Substructures were defined by maching atoms by atomic numbers and bonds by bond order (with aromatic bonds given a distinct bond order), and ring membership.