Performance of HADDOCK and a simple contact-based protein–ligand binding affinity predictor in the D3R Grand Challenge 2
Performance of HADDOCK and a simple contact-based protein-ligand binding affinity predictor in the D3R Grand Challenge 2
Zeynep Kurkcuoglu 0 1 2
Panagiotis I. Koukos 0 1 2
Nevia Citro 0 1 2
Mikael E. Trellet 0 1 2
J. P. G. L. M. Rodrigues 0 1 2
Irina S. Moreira 0 1 2
Jorge Roel-Touris 0 1 2
Adrien S. J. Melquiond 0 1 2
Cunliang Geng 0 1 2
Jörg Schaarschmidt 0 1 2
Li C. Xue 0 1 2
Anna Vangone 0 1 2
A. M. J. J. Bonvin 0 1 2
0 James H. Clark Center, Stanford University , 318 Campus Drive, S210, Stanford, CA 94305 , USA
1 CNC - Center for Neuroscience and Cell Biology, FMUC, Universidade de Coimbra , Rua Larga, Polo I, 1oandar, 3004-517 Coimbra , Portugal
2 Bijvoet Center for Biomolecular Research, Faculty of Science - Chemistry, Utrecht University , Padualaan 8, 3584CH Utrecht , The Netherlands
3 A. M. J. J. Bonvin
We present the performance of HADDOCK, our information-driven docking software, in the second edition of the D3R Grand Challenge. In this blind experiment, participants were requested to predict the structures and binding affinities of complexes between the Farnesoid X nuclear receptor and 102 different ligands. The models obtained in Stage1 with HADDOCK and ligand-specific protocol show an average ligand RMSD of 5.1 Å from the crystal structure. Only 6/35 targets were within 2.5 Å RMSD from the reference, which prompted us to investigate the limiting factors and revise our protocol for Stage2. The choice of the receptor conformation appeared to have the strongest influence on the results. Our Stage2 models were of higher quality (13 out of 35 were within 2.5 Å), with an average RMSD of 4.1 Å. The docking protocol was applied to all 102 ligands to generate poses for binding affinity prediction. We developed a modified version of our contact-based binding affinity predictor PRODIGY, using the number of interatomic contacts classified by their type and the intermolecular electrostatic energy. This simple structure-based binding affinity predictor shows a Kendall's Tau correlation of 0.37 in ranking the ligands (7th best out of 77 methods, 5th/25 groups). Those results were obtained from the average prediction over the top10 poses, irrespective of their similarity/correctness, underscoring the robustness of our simple predictor. This results in an enrichment factor of 2.5 compared to a random predictor for ranking ligands within the top 25%, making it a promising approach to identify lead compounds in virtual screening.
D3R; Drug design data resource; Docking; Binding affinity; Ranking; Intermolecular contacts
-
Zeynep Kurkcuoglu and Panagiotis I. Koukos are the Joint first
authors.
Introduction
Molecular docking is a widely-used tool in computer-aided
drug design to model the three-dimensional (3D) structure
of protein–ligand complexes, study their interactions and
predict their binding affinities [
1
]. Integrated with data from
the experimental techniques like X-ray crystallography and
Nuclear Magnetic Resonance, docking has become a
powerful tool in designing novel therapeutics [
2
]. Docking consists
of two main steps: (i) exploration of protein–ligand
binding poses (sampling) and (ii) identification of biologically
relevant models (scoring). Both steps have their own
challenges such as the flexibility of entities and the accuracy of
the scoring functions. These have been reviewed elsewhere
[
2–4
].
Our integrative, information-driven, f lexible
docking approach HADDOCK [
5, 6
] addresses this structural
modeling problem by using the available experimental
and bioinformatics data to drive the docking process in
combination with a simple but robust scoring function
for ranking. The success of HADDOCK in modeling
protein–protein, protein-nucleic acid and protein–peptide
complexes has been demonstrated numerous times (for a
review, see [7]). HADDOCK is also consistently among
the top scorers and predictors [
8
] in The Critical
Assessment of Predicted Interactions (CAPRI) experiment [
9
],
where participants are expected to predict the 3D structure
of an unknown biomolecular complex, given the sequence
or the structure of the unbound partners.
While HADDOCK has also been used in several
protein–ligand docking studies [
4, 10–16
], no systematic
benchmarking has been reported so far, making the D3R
Grand Challenge 2 a perfect opportunity to assess its
performance for this type of problem for which it was not
originally developed. In this manuscript, we describe our
strategy for predicting the binding poses of FXR ligands
(Stage1), and assessing their binding affinities (Stage2),
while discussing the main lessons learned from the
challenge.
Materials and methods
Data
The target of the D3R Grand Challenge 2 is the Farnesoid
X nuclear receptor (FXR), which is a nuclear hormone
receptor activated by bile acids [
17
]. FXR is highly
expressed in liver, intestines and kidneys, playing an
important role in the regulation of bile acid homeostasis
and cholesterol, lipid and glucose (...truncated)