Table 1.

Datasets with known binding sites used to assess the predictions accuracy

PDB ID	Antibodya	Antigena	Source	Library sizeb
Validation
1JRH	Ab A6	IFNgammaR	(25)	60 × 5
				36 × 6
1BJ1c	VEGF	VEGF Ab	(26)	3 × 5
				2 × 4
Antibody–antigen
1g9M	mAb 17b	gp120	(8)	10 × 14
				1 × 12
1E6J	mAb 13b5	p24	(8)	14 × 14
				2 × 7
1N8Z	Herceptin Fab	Her-2	(34)	5 × 12
1IQD	mAb Bo2C11	Coagulation factor VIII	(35)	27 × 12
Protein–protein
1AVZ	Fyn SH3 domain	Nef	(27)	8 × 10
				10 × 12
1G83	Fyn SH3 domain	SH2	(27)	8 × 10
				10 × 12
1HX1	Bovine Hsc70	Bag chaperone regulator	(28)	8 × 15
Synthesized peptide
1HX1	Bovine Hsc70	Bag chaperone regulator	(28)	1 × 7

^aIn the case of protein–protein datasets, the antibody and the antigen correspond to the target and template proteins, respectively.

^bNumber of sequences × sequence length.

^cIn this experiment, the phage library was screened by VEGF and the mapping is onto the anti-VEGF mAb (see text for details).