Sfcnn: a novel scoring function based on 3D convolutional neural network for accurate and stable protein–ligand affinity prediction

Table 5 Performance of Sfcnn on CASF-2016, CASF-2013, CSAR_HiQ_NRC_set, and Astex_diverse_set datasets after excluding complexes with high structural and chemical similarity to the training set ones

Datasets	R	RMSE	MAE	Size
TM < 0.5, ligand similarity < 0.8
CASF-2016	0.7772	1.4006	1.0931	200
CASF-2013	0.7898	1.5592	1.1882	78
CSAR_HiQ_NRC_set	0.6372	1.8839	1.4630	124
Astex_diverse_set	0.6404	1.3372	1.0505	70
TM < 0.17, ligand similarity < 0.8
CASF-2016	0.8008	1.3731	1.087	170
CASF-2013	0.7863	1.6088	1.2282	65
CSAR_HiQ_NRC_set	0.6353	1.9321	1.5149	75
Astex_diverse_set	0.6356	1.3109	1.0425	53

ISSN: 1471-2105