AutoMap: A tool for analyzing protein–ligand recognition using multiple ligand binding modes

https://doi.org/10.1016/j.jmgm.2013.01.001Get rights and content

Abstract

Prediction of the protein residues most likely to be involved in ligand recognition is of substantial value in structure-based drug design. Considering multiple ligand binding modes is of potential relevance to studying ligand recognition, but is generally ignored by currently available techniques. We have previously presented the site mapping technique, which considers multiple ligand binding modes in its analysis of protein–ligand recognition. AutoMap is a partially automated implementation of our previously developed site mapping procedure. It consists of a series of Perl scripts that utilize the output of molecular docking to generate “site maps” of a protein binding site. AutoMap determines the hydrogen bonding and van der Waals interactions taking place between a target protein and each pose of a ligand ensemble. It tallies these interactions according to the protein residues with which they occur, then normalizes the tallies and maps these to the surface of the protein. The residues involved in interactions are selected according to specific cutoffs. The procedure has been demonstrated to perform well in studying carbohydrate–protein and peptide–antibody recognition. An automated procedure to optimize cutoff selection is demonstrated to rapidly identify the appropriate cutoffs for these previously studied systems. The prediction of key ligand binding residues is compared between AutoMap using automatically optimized cutoffs, AutoMap using a previously selected cutoff, the top ranked pose from docking and the predictions supplied by FTMap. AutoMap using automatically optimized cutoffs is demonstrated to provide improved predictions, compared to other methods, in a set of immunologically relevant test cases. The automated implementation of the site mapping technique provides the opportunity for rapid optimization and deployment of the technique for investigating a broad range of protein–ligand systems.

Highlights

► AutoMap, a new tool for studying ligand recognition, is described. ► Used multiple ligand binding modes from docking to describe recognition. ► Optimization method ensures rapid selection of optimal parameters. ► Significant improvement over other methods for identifying ligand-binding residues. ► Potentially a valuable technique for structure-based drug design.

Introduction

Determining the structural basis of ligand–protein recognition is crucial in structure-based drug design and functional studies. Structural information can be obtained experimentally, using X-ray crystallography and NMR spectroscopy, or by automated molecular docking. While experimental techniques provide valuable insights into ligand–protein recognition, they are often laborious to implement for the rapid determination of protein complexes with distinct ligands, particularly when studying large ligand libraries and highly flexible ligands. Automated molecular docking provides rapid access to binding modes for a large number of protein–ligand systems [1]. However, it is generally accepted that there is no perfect scoring function for molecular docking, and validation studies are usually required to determine the optimal docking algorithm and scoring function for studying a given ligand–protein recognition event [1].

An issue that plagues all of the techniques is their limited consideration of the possibility of multiple binding modes [2], [3]. X-ray crystallography generally provides a single “snapshot” of the ligand binding event, while an ensemble (or ensemble average) of the potential variety of binding states is generally obtained from NMR spectroscopy. A limited number of crystallographic examples where multiple ligand binding modes have been observed with a given target protein have been reported [4], [5], [6], [7], [8]. Molecular docking approaches can provide multiple ligand binding modes, but in practice, these are rarely utilized by the end user in describing ligand binding. Typically, only the top ranked pose (i.e., best scoring) is considered, although some efforts in considering several distinct binding modes to improve docking accuracy have been reported [9].

In our efforts to improve the reliability of binding mode prediction from docking output, we developed the site mapping technique, which utilizes an ensemble of poses obtained from molecular docking to determine the most likely protein residues involved in ligand recognition. The site mapping technique has been demonstrated to perform well in investigating peptide–antibody binding [10], [11], carbohydrate–protein recognition [12], [13], [14] and carbohydrate–peptide mimicry [15], [16]. It forms an integral part of the binding mode determination protocol that we have built around our mapping techniques, which also include the ligand-based epitope and conformational mapping techniques [17].

To increase the accessibility of the site mapping approach, we present AutoMap, a partially automated implementation of the technique. AutoMap requires the target protein and a series of ligand poses obtained from molecular docking as input. The interactions taking place in each pose are determined and processed to generate hydrogen bonding and van der Waals site maps of the protein. The technique is likely to be most useful for examining cases where considering a single (i.e., top ranked) docked pose may result in the wrong prediction or is not sufficiently descriptive of ligand binding, as well as studying ligand binding to shallow binding sites. The former case is exemplified by highly flexible ligands, such as carbohydrates and peptides; the latter case is characteristic of protein–protein interactions, or any protein which utilizes a large, relatively flat surface for ligand binding, such as some lectins [13]. The automated implementation allows rapid deployment of the technique for studying any protein–ligand recognition event. We have also compared the results of AutoMap with that of another mapping method, FTMap, which uses solvent-based probes to identify potential ligand binding sites in proteins.

Section snippets

The AutoMap method

The steps of AutoMap are highlighted in Fig. 1. To start, the protein–ligand system(s) of interest are identified, and the relevant structures obtained. An ensemble of predicted ligand–protein complexes is always required. The interactions taking place in each pose of the ensemble are determined and the hydrogen bonding and van der Waals site maps are generated using the appropriate cutoffs for each interaction type (Fig. 1a). The cutoffs are obtained through the use of the optional (but highly

Results

The site mapping technique has been previously demonstrated for carbohydrate–protein [12], [13], [28] and peptide–antibody [15] systems, and is not re-examined in detail here. When executed as described in the manual, AutoMap can readily generate site maps as in earlier reports [12], [13], [15], [28], [31], [32]. Instead, the results hereby presented will focus on the use of the automated optimization procedure for identifying cutoffs for the validation systems previously studied (Supplementary

Discussion

Approaches which identify key residues involved in ligand binding can be loosely classified as energy-based or empirical. The well-known GRID method [35] is an example of an energy-based approach. It utilizes a water probe to map energetically favourable positions of a potential binding site. Similar methods include Multiple Copy Simultaneous Search (MCSS) [36], [37] and CS-Map [38], which use a variety of small organic probes to determine favourable ligand-binding regions. PP_SITE [39] also

Conclusions

We have developed AutoMap, a fully automated implementation of our previously developed site mapping procedure. AutoMap is generally applicable to studying a wide range of ligand–protein complexes, including the binding of small molecules as well as large flexible molecules such as carbohydrates and peptides. We have also developed a fully automated protocol for optimizing the cumulative sum cutoffs for use with the procedure, which exhaustively examines all possible combinations of hydrogen

Authors’ contributions

MA, PAR and EY conceived the ideas. MA prepared the scripts, carried out the studies and prepared the manuscript. PAR, EY and RLM edited the manuscript. All authors read and approved the final manuscript.

Acknowledgements

This research was supported by a small grant from the Faculty of Pharmacy and Pharmaceutical Sciences, Monash University, to E.Y. M.A. was a recipient of the Monash University Postgraduate Publication Award (PPA). P.A.R. is the Sir Zelman Cowen Senior Research Fellow (Sir Zelman Cowen Fellowship Fund, Burnet Institute). The authors gratefully acknowledge the contribution to this work of the Victorian Operational Infrastructure Support Program received by the Burnet Institute. The funders had no

References (51)

  • A.S. Goré et al.

    Structure of the adenosine A2A receptor in complex with ZM241385 and the xanthines XAC and caffeine

    Structure

    (2011)
  • R.E. Watkins et al.

    The human nuclear xenobiotic receptor PXR: structural determinants of directed promiscuity

    Science

    (2001)
  • P. Källblad et al.

    Assessment of multiple binding modes in ligand–protein docking

    Journal of Medicinal Chemistry

    (2004)
  • E. Yuriev et al.

    Docking of combinatorial peptide libraries into a broadly cross-reactive human IgM

    Journal of Molecular Recognition

    (2001)
  • E. Yuriev et al.

    Recognition of IgG-derived peptides by a human IgM with an unusual combining site

    Scandinavian Journal of Immunology

    (2002)
  • M. Agostino et al.

    A computational approach for exploring carbohydrate recognition by lectins in innate immunity

    Frontiers in Immunology

    (2011)
  • J. Milland et al.

    Carbohydrate residues downstream of the terminal Galα(1,3)Gal epitope modulate the specificity of xenoreactive antibodies

    Immunology and Cell Biology

    (2007)
  • M. Agostino et al.

    Peptide inhibitors of xenoreactive antibodies mimic the interaction profile of the native carbohydrate antigens

    Biopolymers

    (2011)
  • E. Yuriev et al.

    Antibody–ligand docking: insights into peptide–carbohydrate mimicry

    Molecular Simulation

    (2008)
  • M. Agostino et al.

    Identification of preferred carbohydrate binding modes in xenoreactive antibodies by combining conformational filters and binding site maps

    Glycobiology

    (2010)
  • Worldwide Protein Data Bank

    Protein Data Bank Contents Guide: Atomic Coordinate Entry Format Description Version 3.30

    (2011)
  • A.C. Wallace et al.

    LIGPLOT: a program to generate schematic diagrams of protein–ligand interactions

    Protein Engineering

    (1995)
  • D.K. Chalmers et al.

    Silico – A Perl Molecular Modelling Toolkit

    (2011)
  • R.A. Friesner et al.

    Glide: a new approach for rapid, accurate docking and scoring. 1. Method and assessment of docking accuracy

    Journal of Medicinal Chemistry

    (2004)
  • Schrödinger, LLC. Glide, Version 5.6, New York, NY,...
  • Cited by (0)

    View full text