HADDOCK/PRODIGY scoring protocol
HADDOCK2.2webserver/devel version of PRODIGY
As for stage1
The ranking prediction was performed using the contact-based approach introduced in "Vangone and Bonvin eLife, 2015", and developed into the PRODIGY web-server (Xue et al, Bioinformatics 2016), properly re-adapted for protein-ligand complexes. The scoring is based on the average of the top5 poses. For each pose, the number of atomic contacts between the protein and the ligand within the distance threshold of 10.5 Angstrom was calculated. Contacts were further classified according to the atom involved in the interaction (C= carbon, O= oxygen, N= nitrogen, X= all the other atoms). The contact-based ranking function is defined as: 0.343794*Eelec - 0.037597*CC + 0.138738*NN + 0.160043*OO - 3.088861*XX + 187.011384, where CC, NN, OO and XX are the atomic-contacts classified by atom-type and Eelec is the electrostatic energy calculated by HADDOCK. The contacts have been calculated with a in-house version of PRODIGY for ligands, currently in development; the electrostatic energy is calculated using the OPLS united atom force field parameters (Jorgensen, W. L. & Tirado-Rives. J. Am. Chem. Soc. 1988) for non-bonded atoms, using a 8.5 Angstroms cut-off with a shifting function for the electrostatic energy. A dielectric constant of 10 is used..
HADDOCK2.2 protein-ligand protocol
Omega TK 2.6.4, MMTSB Tool Set,ProFit 3.1,PRODRG,HADDOCK2.2 webserver,PyMOL v1.8.4
Assumed pH neutral
Protein charges from OPLS force field
Ligand parameters and charges from PRODRG
Based on ligand positioning in the identified templates from Stage1, the binding pocket was defined as all residues within 4 Angstrom to their respective ligand in the various receptors. For the second stage, we identified the ligand that is most similar to FXR1-36 for targets FXR37-102 based on Tanimoto distance. The corresponding receptor was used as a template for the protein during the docking. For targets FXR1-36 the structures provided by the organisers at the end of stage 1 were used.
For the ligands, we converted the SMILES strings provided by the organisers, to 3D structures using Omega while generating conformers using Omega torsional sampling at the same time. The maximum number of conformers generated per ligand was capped to 100. The clustering of conformers is performed by MMTSB tools using the following settings: jclust hierarchical clustering with maximum number of clusters set to 10 and minimum number of elements is set to 4 per cluster. Representative structure from each cluster is selected while up sampling the major cluster, an ensemble is formed by representatives for each ligand.
Default HADDOCK2.2 server setting except for the following changes
it0 sampling (rigid body docking): 10000 models
it1 sampling (flexible refinement): 400 models
delenph = False
inter_rigid = 0.001
tadinit2_t = 500
tadfinal2_t = 50
tadinit3_t = 500
tadfinal3_t = 50
initiosteps = 0
cool1_steps = 0
w_vdw_0 =0.0
amb = ExtStageConstants (firstit = 0, lastit = 0,)
The docking was performed using the HADDOCK2.2 web server (van Zundert et al. J. Mol. Biol. 2015) using default parameters except for those specified above. Ensembles of 5 receptor structures and various ligand conformations were provided as input for ensemble docking.
Two restraints files were provided to the server to guide the docking: 1) an ambig.tbl file in which the ligand and all residues in the binding pocket were defined as active - this file was only used in it0 and 50% of the restraints were randomly deleted for each docking trial 2) an unambig.tbl file in which only the ligand is defined as active and the protein binding pocket as passive (meaning that no energy penalty is generated if a binding pocket residue does not contact the ligand). This second file was used at all stages of the docking protocol.
The top 5 poses from the HADDOCK2.2-it1 stage were selected for submission.
The scoring function used for ranking the poses is the standard HADDOCK score for the flexible refinement (It1) which is defined as: HADDOCK-score = 1.0*Evdw + 1.0*Eelec + 1.0*Edesol - 0.01*BSA, where BSA is the buried surface area in Angstrom**2, Edesol an empirical desolvation energy term (Fernandez-Recio et al. J. Mol. Biol. 2004). The intermolecular energies are calculated using the OPLS united atom force field parameters (Jorgensen, W. L. & Tirado-Rives. J. Am. Chem. Soc. 1988) for non-bonded atoms, using a 8.5 cut-off with a shifting function for the electrostatic energy and switching function between 6.5 and 8.5 for the van der Waals energy. For the electrostatics energy, a dielectric constant of 10 is used.