Predicting Ligand Binding Affinity with Alchemical Free Energy Methods in a Polar Model Binding Site

Graduate Group in Biophysics, University of California-San Francisco, San Francisco, CA 94158-2518, USA
Journal of Molecular Biology (Impact Factor: 4.33). 12/2009; 394(4):747-763. DOI: 10.1016/j.jmb.2009.09.049


We present a combined experimental and modeling study of organic ligand molecules binding to a slightly polar engineered cavity site in T4 lysozyme (L99A/M102Q). For modeling, we computed alchemical absolute binding free energies. These were blind tests performed prospectively on 13 diverse, previously untested candidate ligand molecules. We predicted that eight compounds would bind to the cavity and five would not; 11 of 13 predictions were correct at this level. The RMS error to the measurable absolute binding energies was 1.8 kcal/mol. In addition, we computed “relative” binding free energies for six phenol derivatives starting from two known ligands: phenol and catechol. The average RMS error in the relative free energy prediction was 2.5 kcal/mol (phenol) and 1.1 kcal/mol (catechol). To understand these results at atomic resolution, we obtained x-ray co-complex structures for nine of the diverse ligands and for all six phenol analogs. The average RMSD of the predicted pose to the experiment was 2.0 Å (diverse set), 1.8 Å (phenol-derived predictions), and 1.2 Å (catechol-derived predictions). We found that predicting accurate affinities and rank-orderings required near-native starting orientations of the ligand in the binding site. Unanticipated binding modes, multiple ligand binding, and protein conformational change all proved challenging for the free energy methods. We believe that these results can help guide future improvements in physics-based absolute binding free energy methods.


Available from: Ken A Dill
  • Source
    • "They correspond to residues that contribute significantly in the binding of lapatinib but participate minimally in the association of ATP (Figure 7A). While realistic quantitative predictions of these contributions is difficult currently owing to deficiencies in force fields and lack of exhaustive sampling [24], our approach begins to distill out the differing contributions within the different subpopulations thus highlighting the importance of accounting for this significant heterogeneity of conformations. The putative mutants appear to be confined to regions adjacent to the binding pockets of the enzymes (Figure 3 and Figure 4). "
    [Show abstract] [Hide abstract]
    ABSTRACT: The pressure exerted by drugs targeted to a protein in any therapy inevitably leads to the emergence of drug resistance. One major mechanism of resistance involves the mutation of key residues in the target protein. Drugs that competitively replace a natural substrate are often made ineffective by mutations that reduce the drug's affinity relative to that of the natural substrate. Hence atomic level understanding of the mechanisms that underlie this behavior is of utmost importance in efforts to design new drugs that can target such mutant proteins. Methods that can predict these mutations before they appear in clinic would be a major advance in the selection of the appropriate treatment strategy in patients. The present computational approach aims to model this emergence in EGFR and ErbB2 after treatment with the drug lapatinib, by investigating the structural, dynamic and energetic effects on these kinases when bound to the natural substrate ATP and to lapatinib. The study reveals binding modes and subpopulations that are presumably normally cryptic and these have been analyzed extensively here with respect to sites that are predicted to be hotspots for resisting mutations. These positions are compared in the context of currently available data from laboratory-based experiments and mechanistic details, at the atomistic level, of the origin of resistance are developed. The prediction of novel mutations, if validated by their emergence in the clinic, will make these methods as a powerful predictive tool which can be used in the design of new kinase inhibitors.
    PLoS ONE 10/2013; 8(10):e77054. DOI:10.1371/journal.pone.0077054 · 3.23 Impact Factor
  • Source
    • "While HTS has illuminated docking false negatives [56]; docking has correspondingly illuminated false negatives from HTS[3]. Ever more frequently, docking predictions are tested by subsequent x-ray crystallographic structures, often confirming the predicted geometries of the docked complex[7], [14], [59]–[65]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Molecular docking remains an important tool for structure-based screening to find new ligands and chemical probes. As docking ambitions grow to include new scoring function terms, and to address ever more targets, the reliability and extendability of the orientation sampling, and the throughput of the method, become pressing. Here we explore sampling techniques that eliminate stochastic behavior in DOCK3.6, allowing us to optimize the method for regularly variable sampling of orientations. This also enabled a focused effort to optimize the code for efficiency, with a three-fold increase in the speed of the program. This, in turn, facilitated extensive testing of the method on the 102 targets, 22,805 ligands and 1,411,214 decoys of the Directory of Useful Decoys - Enhanced (DUD-E) benchmarking set, at multiple levels of sampling. Encouragingly, we observe that as sampling increases from 50 to 500 to 2000 to 5000 to 20000 molecular orientations in the binding site (and so from about 1×10(10) to 4×10(10) to 1×10(11) to 2×10(11) to 5×10(11) mean atoms scored per target, since multiple conformations are sampled per orientation), the enrichment of ligands over decoys monotonically increases for most DUD-E targets. Meanwhile, including internal electrostatics in the evaluation ligand conformational energies, and restricting aromatic hydroxyls to low energy rotamers, further improved enrichment values. Several of the strategies used here to improve the efficiency of the code are broadly applicable in the field.
    PLoS ONE 10/2013; 8(10):e75992. DOI:10.1371/journal.pone.0075992 · 3.23 Impact Factor
  • Source
    • "Because of their simplicity, docking against these model cavities has revealed particular errors in our scoring functions and our representation of molecular properties, most often by the misprediction of molecules, which in these simple sites are often illuminating. Examples include the importance of using higher-level partial atomic charges for ligands [20], the challenges posed by decoy molecules when van der Waals repulsion terms are softened [24], the need to account for strain energy when modeling receptor flexibility [25], the trade-offs between optimizing geometric fidelity and ligand discovery [26], the consequences of neglecting ordered and especially bridging waters in the docking calculations [27], the challenges of correctly balancing van der Waals and electrostatic interaction terms in docking [21], and the opportunities and challenges for even the highest level of theory to predict binding affinities in these simple sites [28]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: A key challenge in structure-based discovery is accounting for modulation of protein-ligand interactions by ordered and bulk solvent. To investigate this, we compared ligand binding to a buried cavity in Cytochrome c Peroxidase (CcP), where affinity is dominated by a single ionic interaction, versus a cavity variant partly opened to solvent by loop deletion. This opening had unexpected effects on ligand orientation, affinity, and ordered water structure. Some ligands lost over ten-fold in affinity and reoriented in the cavity, while others retained their geometries, formed new interactions with water networks, and improved affinity. To test our ability to discover new ligands against this opened site prospectively, a 534,000 fragment library was docked against the open cavity using two models of ligand solvation. Using an older solvation model that prioritized many neutral molecules, three such uncharged docking hits were tested, none of which was observed to bind; these molecules were not highly ranked by the new, context-dependent solvation score. Using this new method, another 15 highly-ranked molecules were tested for binding. In contrast to the previous result, 14 of these bound detectably, with affinities ranging from 8 µM to 2 mM. In crystal structures, four of these new ligands superposed well with the docking predictions but two did not, reflecting unanticipated interactions with newly ordered waters molecules. Comparing recognition between this open cavity and its buried analog begins to isolate the roles of ordered solvent in a system that lends itself readily to prospective testing and that may be broadly useful to the community.
    PLoS ONE 07/2013; 8(7):e69153. DOI:10.1371/journal.pone.0069153 · 3.23 Impact Factor
Show more