Table 1. Performance comparison of COACH-D and COACH-D 2.0 on 20 multimeric structures. The structures are randomly selected from Q-BioLiP by enforcing the following criteria: (1) exclusion of structures with >5 chains or <5 binding residues; (2) pairwise sequence similarity ≤ 30%; (3) ligands are small molecules; (4) the total number of residues in the protein structure is less than 2000.
a results of COACH-D , b results of COACH-D 2.0.

No. PDB ID Ligand ID Chain MCCa MCCb Time(h)a Time(h)b
1 1a6w NIP 2 0.673 0.884 7.9 0.7
2 4m20 COA 4 0.655 0.613 7.8 1.3
3 4m8y KGQ 2 0.681 0.815 3.1 0.9
4 4m99 ACO 3 0.546 0.556 8.9 1.0
5 4n5f FDA 2 0.706 0.680 10.4 1.4
6 4p86 5GP 4 -0.007 0.168 10.3 1.0
7 6zth NAD 4 -0.003 0.710 21.9 0.8
8 5suv COA 3 0.842 0.861 5.5 0.8
9 5xuh COA 3 0.739 0.806 4.7 0.5
10 6a03 2BA 2 0.793 0.883 4.7 0.3
11 6dft NAD 4 0.695 0.695 17.1 1.4
12 6p25 CPL 2 -0.004 0.922 19.4 1.3
13 6w3z NAD 4 0.766 0.777 19.7 1.3
14 6xb3 9BG 2 0.508 0.671 7.7 0.1
15 7b6b SZQ 2 0.069 0.000 11.7 0.9
16 7bxr F9X 2 0.807 0.705 12.7 1.6
17 6o6y LQJ 2 0.584 0.974 10.7 0.4
18 8dor FMN 2 0.974 0.974 5.3 0.9
19 8e77 ULP 2 0.869 0.860 11.9 1.5
20 8eg6 US9 4 0.633 0.631 7.1 1.3
21 Average -- -- 0.576 0.709 10.4 1.0