Docking with ClusPro, HADDOCK, and Rosie

It should be noted that the methodology used is not unique to ICM, as such we have performed three docking experiments with commonly available online docking servers. Docking was performed with three common servers: ClusPro^42–44, HADDOCK^45,46, and Rosie^47,48 with the PP FabF. The results are reported in Table S8, comparing the top three structures generated from the servers. Models created by ICM, ClusPro, HADDOCK, and Rosie were compared to the crosslinked crystal structures through two methods. The first was a full alignment in Pymol, this yielded good values but often global changes in the PP and AcpP either upon crosslinking or due to differences in substrate or crystallization conditions appeared to have altered the backbone distant from the interface. To specifically look at the interfaces between the two enzymes the atoms within 5 Å of the interface were selected for both the docked and crystal structures. The two interfaces were then superimposed upon one another using the Pymol super command and the value was reported without deletion. Pymol was used in all structural visualizations. For the comparison study using online servers the input files used in all docking, post solvation, and minimization were collected. For HADDOCK docking both the ICM prepared files and the raw PDB files (2FAD and 2GFW) the inputs were loaded into the server and active residues were defined as 65 and 616 on FabF and 35 and 39 on AcpP, the same focus residues in the ICM study. Passive residues were defined within 6.5 Å of the active residues as was suggested by the program. As the study is meant to sample the most accessible components of the method the “EASY” access level account was used. Meaning all parameter settings were set to default. The results were ranked using the standard energy scoring metric in HADDOCK and benchmarked by alignment to the crosslinked structure. The Cluspro docking was the simplest method performed for this study; the same input structures were used as the HADDOCK. After loading in the structures the attraction residues were set to be the same as in the HADDOCK and ICM jobs. Both the electrostatic and balanced scoring functions are presented. The Rosie server was run with the Docking2 refinement utility, as this utility requires an input structure the Cluspro best ranked “Electrostatic scoring” docked file was used. We felt this represented a second refinement step which could easily be taken by other groups after the rapid Cluspro docking. However, we would advise caution when performing refinements of ClusPro docking. Great care should be taken to ensure that the starting structures are an accurate starting structure and that Rosie is not optimizing an incorrect interface. We feel this demonstrates the ability of many utilities to re-create AcpP partner interfaces. As well it demonstrates the importance of caution when evaluating docked structures from Cluspro. HADDOCK performed extremely well when supplied with properly prepared starting structures, but we feel this stands as additional evidence that preparation of the proteins to re-create a solvated structure is important. As like ICM HADDOCK performed poorly when provided with structures straight from the PDB.