Table 2.
Assessment of SiteLight’s performance on the dataset presented in Table 1
| No. | Fragments | Surface residues | Interface residues (%) | Predicted residues | Matches (selected) | Highest solution | Best solution | N100 |
|---|---|---|---|---|---|---|---|---|
| 1A | 2 | 104 | 21 (20) | 13 | 832 (3) | 1 (68) | 1 (68) | 1 (68) |
| 1B | 2 | 104 | 21 (20) | 12 | 306 (12) | 1 (28) | 9 (83) | 21 (100) |
| 2 | 1 | 58 | 12 (20) | 6 | 990 (3) | 1 (100) | 1 (100) | 1 (100) |
| 4 | 4 | 98 | 14 (14) | 3 | 1692 (3) | 1 (27) | 1 (27) | 65 (90) |
| 5 | 4 | 55 | 12 (21) | 1 | 988 (4) | none | 1 (14) | 50 (75) |
| 6A | 2 | 48 | 11 (22) | 3 | 912 (10) | 1 (50) | 1 (50) | 11 (100) |
| 6B | 2 | 48 | 11 (22) | 10 | 576 (7) | 1 (100) | 1 (100) | 1 (100) |
| 7 | 13 | 305 | 44 (14) | 17 | 19500 (29) | 1 (25) | 27 (66) | 27 (66) |
| 8 | 6 | 55 | 18 (32) | 5 | 624 (3) | 1 (40) | 1 (40) | 65 (60) |
| 9A | 2 | 52 | 12 (23) | 3 | 312 (6) | 1 (100) | 1 (100) | 1 (100) |
| 9B | 2 | 52 | 12 (23) | 8 | 572 (6) | 1 (100) | 1 (100) | 1 (100) |
| 9C | 2 | 52 | 12 (23) | 6 | 624 (6) | 1 (100) | 1 (100) | 1 (100) |
| 10 | 4 | 214 | 21 (9) | 3 | 6400 (11) | 7 (27) | 7 (27) | 29 (70) |
| 11 | 2 | 56 | 11 (19) | 7 | 658 (16) | 1 (40) | 8 (80) | 20 (100) |
| Dimers | ||||||||
| 12 | 5 | 70 | 14 (20) | 6 | 396 (5) | 2 (66) | 2 (66) | 28 (100) |
| 13A | 7 | 205 | 14 (6) | 8 | 3800 (31) | 3 (40) | 6 (75) | 6 (75) |
| 13B | 7 | 205 | 14 (6) | 5 | 1200 (5) | none | 5 (15) | 88 (53) |
| 14 | 3 | 156 | 8 (5) | 0 | 2556 (25) | none | none | 63 (36) |
| Peptide–Protein | ||||||||
| 15 | 6 | 104 | 22 (21) | 8 | 7650 (57) | 7 (60) | 7 (60) | 100 (83) |
| Antigen–Antibody | ||||||||
| 16A | 4 | 406 | 13 (3) | 5 | 2067 (27) | 7 (28) | 7 (28) | 43 (33) |
| 16B | 4 | 406 | 13 (3) | 4 | 1510 (24) | 4 (33) | 4 (33) | 70 (40) |
| 17A | 4 | 205 | 18 (8) | 12 | 1188 (63) | 1 (100) | 1 (100) | 1 (100) |
| 17B | 4 | 205 | 18 (8) | 4 | 985 (22) | none | 3 (16) | 35 (100) |
| 17C | 4 | 205 | 18 (8) | 6 | 6138 (24) | 2 (83) | 2 (83) | 37 (100) |
| 18A | 5 | 94 | 15 (15) | 8 | 2093 (6) | 1 (85) | 1 (85) | 11 (100) |
| 18B | 5 | 94 | 15 (15) | 11 | 637 (10) | 1 (85) | 4 (100) | 4 (100) |
| 18C | 5 | 94 | 15 (15) | 9 | 91 (6) | 1 (33) | 4 (100) | 4 (100) |
| Antigen–Fc | ||||||||
| 19A | 2 | 43 | 11 (25) | 8 | 86 (2) | 2 (38) | 2 (38) | 75 (81) |
| 19B | 1 | 54 | 13 (24) | 12 | 108 (2) | 1 (60) | 1 (60) | 21 (68) |
| 19C | 4 | 223 | 9 (4) | 0 | 406 (3) | none | none | 98 (26) |
| 19D | 5 | 90 | 26 (28) | 12 | 1600 (3) | 2 (57) | 2 (57) | 91 (70) |
| 20 | 3 | 197 | 15 (7) | 2 | 1940 (17) | none | 14 (18) | 63 (41) |
| Hapten–Antibody | ||||||||
| 21 | 4 | 408 | 9 (2) | 4 | 1548 (32) | none | 22 (14) | 85 (25) |
| 22A | 3 | 378 | 9 (2) | 0 | 7260 (22) | none | none | 73 (71) |
| 22B | 3 | 378 | 9 (2) | 4 | 6897 (143) | 1 (60) | 1 (60) | 1 (60) |
| 22C | 3 | 378 | 9 (2) | 1 | 2541 (26) | none | none | 81 (60) |
| 22D | 3 | 378 | 9 (2) | 1 | 2904 (24) | none | none | 59 (100) |
Unlike in Table 2▶ in the Supplemental Material, here the entire corresponding phage display library is used. All of the categories presented in this table refer to 25% coverage of the surface. The number of potential matches used to predict interface residues is determined such that the number of predicted residues will equal 25% of the surface residues.
Fragments, the number of contiguous sequences in the spatially defined interface of the complex; surface residues, the number of residues exposed to the solvent; interface residues, Template residues that are proximate to the Target in space. The interface percent is calculated with regard to surface residues as the 100%; predicted residues, the number of interface residues that were correctly predicted by SiteLight; matches, the number of potential matches. This number equals the number of surface patches multiplied by the number of peptides in the library. The number of the obtained matches for binding site prediction is indicated in parentheses; highest solution, the rank of the first solution that overlaps by 25% or more with the interface out of the obtained matches. The percent of residues that overlap with the interface is indicated in parentheses; best solution, the rank of the match that overlaps with the interface to the largest extent; N100, the rank of a match that overlaps with the interface to the largest extent out of the first 100 solutions. See Halperin et al. (2002). The results refer to 25% coverage of the surface.
Example: in case No. 1A there are 104 surface residues. The interface is composed of 21 residues that are 20% of the surface residues. The interface is made up of two sequential regions that are proximate in space to one another and to the Target. There are 104 potential matches that are tested by SiteLight. Using the threshold that we allow, no more than 25% of the surface of the molecule is to be covered by the matches (otherwise, the prediction of a binding site is too diffuse). Two matches (out of the possible 104) were obtained. Twenty-five percent of the surface is 26 residues. The two matches contain 17 interface residues (out of a possible 21 identified in the interface). Sixty-eight percent of the residues in the highest ranking match (solution no. 1) are interface residues. The second solution overlaps with the interface to a lesser extent. Therefore, in this example, the highest ranking solution is also the best solution with the largest interface coverage. Ninety-two percent of the residues of the 56th ranked match are interface residues. It is the highest overlap with the interface in the first 100 ranked matches.