Table 2. Performance comparison of different refinement methods in terms of mean improvement/percentage of improved cases in ligand RMSD (L-RMSD), interface RMSD (I-RMSD), fraction of native contact (Fnat), and MolProbity score (MolP).
| Hetero-complex refinement | |||||
|---|---|---|---|---|---|
| Test set | Method | −ΔL-RMSD (Å) | −ΔI-RMSD (Å) | ΔFnat (%) | −ΔMolP |
| ZDOCK benchmark set (677 models) | GalaxyRefineComplex | 0.68/82% (0.06/54%) <0.11/56%> | 0.37/85% (0.15/59%) <0.12/64%> | 10.3/91% (7.7/81%) <6.9/87%> | 1.41/100% (1.33/100%) <1.32/100%> |
| RosettaDock | 0.40/56% (−3.73/29%) <−3.11/25%> | 0.30/61% (−1.61/32%) <−1.33/26%> | 2.2/58% (−8.4/30%) <−7.9/33%> | 0.35/95% (0.33/95%) <0.32/95%> | |
| FiberDock1 | –/– (−0.59/41%) | –/– (−0.19/44%) | –/– (−2.2/34%) | –/– (−0.02/44%) | |
| CAPRI set (34 models) | GalaxyRefineComplex | 0.48/74% (0.20/65%) <0.15/60%> | 0.47/90% (0.02/68%) <0.12/60%> | 11.0/82% (7.6/61%) <7.8/82%> | 0.95/100% (0.85/98%) <0.84/100%> |
| RosettaDock | −2.38/48% (−3.95/30%) <−3.80/30%> | −3.38/45% (−3.99/22%) <−4.29/20%> | 7.2/63% (2.6/46%) <2.2/52%> | 0.16/92% (0.13/81%) <0.13/84%> | |
| FiberDock1 | –/– (−0.19/41%) | –/– (0.06/40%) | –/– (4.8/70%) | –/– (0.00/0%) | |
|
Homo-complex refinement | |||||
| PISA benchmark set (445 models) | GalaxyRefineComplex | 1.30/89% (0.78/73%) <0.63/76%> | 0.63/86% (0.37/72%) <0.32/73%> | 8.5/73% (6.1/64%) <4.5/64%> | 1.36/100% (1.25/100%) <1.25/100%> |
| RosettaDock | 3.49/84% (0.82/73%) <0.38/69%> | 1.74/86% (0.17/73%) <−0.08/67%> | 17.1/74% (6.6/61%) <1.9/55%> | 0.66/99% (0.57/99%) <0.56/99%> | |
| SymmRef1 | –/– (2.90/80%) | –/– (1.14/79%) | –/– (14.6/64%) | –/– (−2.55/0%) | |
| CAPRI set (60 models) | GalaxyRefineComplex | 0.08/65% (−0.05/60%) <−0.04/60%> | 0.07/74% (0.02/71%) <0.01/65%> | 2.6/70% (0.8/57%) <0.9/57%> | 0.78/89% (0.73/83%) <0.72/83%> |
| RosettaDock | −1.22/38% (−3.31/22%) <−3.97/20%> | −0.86/52% (−1.71/30%) <−2.19/25%> | −5.9/28% (−9.2/18%) <−11.6/15%> | 0.06/74% (0.03/70%) <0.03/73%> | |
| SymmRef1 | –/– (−0.43/38%) | –/– (−0.20/37%) | –/– (−4.3/39%) | –/– (0.00/0%) | |
Results for the best of 10 refined models are presented, and those for model 1’s and the mean of the 10 models are shown in parentheses and pointy brackets, respectively.
1Data for the 10 models and the mean of the 10 models are not provided for FiberDock and SymmRef because they generate only single models.