Exhaustive search and solvated interaction energy (SIE) for virtual screening and affinity prediction

Solvated interaction energy: from small-molecule to antibody drug design

Scoring functions are ubiquitous in structure-based drug design as an aid to predicting binding modes and estimating binding affinities. Ideally, a scoring function should be broadly applicable, obviating the need to recalibrate and refit its parameters for every new target and class of ligands. Traditionally, drugs have been small molecules, but in recent years biologics, particularly antibodies, have become an increasingly important if not dominant class of therapeutics. This makes the goal of having a transferable scoring function, i.e., one that spans the range of small-molecule to protein ligands, even more challenging. One such broadly applicable scoring function is the Solvated Interaction Energy (SIE), which has been developed and applied in our lab for the last 15 years, leading to several important applications. This physics-based method arose from efforts to understand the physics governing binding events, with particular care given to the role played by solvation. SIE has been used by us and many independent labs worldwide for virtual screening and discovery of novel small-molecule binders or optimization of known drugs. Moreover, without any retraining, it is found to be transferrable to predictions of antibody-antigen relative binding affinities and as accurate as functions trained on protein-protein binding affinities. SIE has been incorporated in conjunction with other scoring functions into ADAPT (Assisted Design of Antibody and Protein Therapeutics), our platform for affinity modulation of antibodies. Application of ADAPT resulted in the optimization of several antibodies with 10-to-100-fold improvements in binding affinity. Further applications included broadening the specificity of a single-domain antibody to be cross-reactive with virus variants of both SARS-CoV-1 and SARS-CoV-2, and the design of safer antibodies by engineering of a pH switch to make them more selective towards acidic tumors while sparing normal tissues at physiological pH.

Accurate prediction of relative binding affinities of a series of HIV-1 protease inhibitors using semi-empirical quantum mechanical charge

A major challenge in computer-aided drug design is the accurate estimation of ligand binding affinity. Here, a new approach that combines the adaptive steered molecular dynamics (ASMD) and partial atomic charges calculated by semi-empirical quantum mechanics (SQMPC), namely ASMD-SQMPC, is suggested to predict the ligand binding affinities, with 24 HIV-1 protease inhibitors as testing examples. In the ASMD-SQMPC, the relative binding free energy (ΔG) is reflected by the average maximum potential of mean force (<PMF>_max ) between bound and unbound states. The correlation coefficient (R² ) between the <PMF>_max and experimentally determined ΔG is 0.86, showing a significant improvement compared with the conventional ASMD (R² = 0.52). Therefore, this study provides an efficient approach to predict the relative ΔG and reveals the significance of precise partial atomic charges in the theoretical simulations.

Improving the accuracy of predicting protein-ligand binding-free energy with semiempirical quantum chemistry charge

Aim: It is a challenge to predict binding-free energy (ΔG) accurately. Methodology/results: For accurate ΔG prediction, a new strategy combining solvated interaction energy (SIE) or molecular mechanics/generalized Born surface area (MM/GBSA) approach with the Coulson charge of both protein and ligand calculated by semiempirical quantum mechanics (SQM), named SIE-SQMPC or MM/GBSA-SQMPC approach, was developed and tested on 50 protein-ligand complexes. Both approaches achieved higher correlation (R ²) between experimental and predicted ΔG than that with Amber-ff03 charge, even for ligands with highly different scaffolds. But, SIE-SQMPC is computationally much faster than MM/GBSA-SQMPC. Conclusion: SIE-SQMPC provided an effective alternative to predict ΔG of protein-ligand binding (R ² = 0.66-0.94 for SIE-AM1; R ² = 0.59-0.98 for SIE-PM7), which has the potential of high-throughput processing for molecular docking and drug design.

Improving Docking Performance Using Negative Image-Based Rescoring

Despite the large computational costs of molecular docking, the default scoring functions are often unable to recognize the active hits from the inactive molecules in large-scale virtual screening experiments. Thus, even though a correct binding pose might be sampled during the docking, the active compound or its biologically relevant pose is not necessarily given high enough score to arouse the attention. Various rescoring and post-processing approaches have emerged for improving the docking performance. Here, it is shown that the very early enrichment (number of actives scored higher than 1% of the highest ranked decoys) can be improved on average 2.5-fold or even 8.7-fold by comparing the docking-based ligand conformers directly against the target protein's cavity shape and electrostatics. The similarity comparison of the conformers is performed without geometry optimization against the negative image of the target protein's ligand-binding cavity using the negative image-based (NIB) screening protocol. The viability of the NIB rescoring or the R-NiB, pioneered in this study, was tested with 11 target proteins using benchmark libraries. By focusing on the shape/electrostatics complementarity of the ligand-receptor association, the R-NiB is able to improve the early enrichment of docking essentially without adding to the computing cost. By implementing consensus scoring, in which the R-NiB and the original docking scoring are weighted for optimal outcome, the early enrichment is improved to a level that facilitates effective drug discovery. Moreover, the use of equal weight from the original docking scoring and the R-NiB scoring improves the yield in most cases.

Binding pose and affinity prediction in the 2016 D3R Grand Challenge 2 using the Wilma-SIE method

The Farnesoid X receptor (FXR) exhibits significant backbone movement in response to the binding of various ligands and can be a challenge for pose prediction algorithms. As part of the D3R Grand Challenge 2, we tested Wilma-SIE, a rigid-protein docking method, on a set of 36 FXR ligands for which the crystal structures had originally been blinded. These ligands covered several classes of compounds. To overcome the rigid protein limitations of the method, we used an ensemble of publicly available structures for FXR from the PDB. The use of the ensemble allowed Wilma-SIE to predict poses with average and median RMSDs of 2.3 and 1.4 Å, respectively. It was quite clear, however, that had we used a single structure for the receptor the success rate would have been much lower. The most successful predictions were obtained on chemical classes for which one or more crystal structures of the receptor bound to a molecule of the same class was available. In the absence of a crystal structure for the class, observing a consensus binding mode for the ligands of the class using one or more receptor structures of other classes seemed to be indicative of a reasonable pose prediction. Affinity prediction proved to be more challenging with generally poor correlation with experimental IC50s (Kendall tau ~ 0.3). Even when the 36 crystal structures were used the accuracy of the predicted affinities was not appreciably improved. A possible cause of difficulty is the internal energy strain arising from conformational differences in the receptor across complexes, which may need to be properly estimated and incorporated into the SIE scoring function.

Assessment of Solvated Interaction Energy Function for Ranking Antibody-Antigen Binding Affinities

Affinity modulation of antibodies and antibody fragments of therapeutic value is often required in order to improve their clinical efficacies. Virtual affinity maturation has the potential to quickly focus on the critical hotspot residues without the combinatorial explosion problem of conventional display and library approaches. However, this requires a binding affinity scoring function that is capable of ranking single-point mutations of a starting antibody. We focus here on assessing the solvated interaction energy (SIE) function that was originally developed for and is widely applied to scoring of protein-ligand binding affinities. To this end, we assembled a structure-function data set called Single-Point Mutant Antibody Binding (SiPMAB) comprising several antibody-antigen systems suitable for this assessment, i.e., based on high-resolution crystal structures for the parent antibodies and coupled with high-quality binding affinity measurements for sets of single-point antibody mutants in each system. Using this data set, we tested the SIE function with several mutation protocols based on the popular methods SCWRL, Rosetta, and FoldX. We found that the SIE function coupled with a protocol limited to sampling only the mutated side chain can reasonably predict relative binding affinities with a Spearman rank-order correlation coefficient of about 0.6, outperforming more aggressive sampling protocols. Importantly, this performance is maintained for each of the seven system-specific component subsets as well as for other relevant subsets including non-alanine and charge-altering mutations. The transferability and enrichment in affinity-improving mutants can be further enhanced using consensus ranking over multiple methods, including the SIE, Talaris, and FOLDEF energy functions. The knowledge gained from this study can lead to successful prospective applications of virtual affinity maturation.

CSAR 2014: A Benchmark Exercise Using Unpublished Data from Pharma

The 2014 CSAR Benchmark Exercise was the last community-wide exercise that was conducted by the group at the University of Michigan, Ann Arbor. For this event, GlaxoSmithKline (GSK) donated unpublished crystal structures and affinity data from in-house projects. Three targets were used: tRNA (m1G37) methyltransferase (TrmD), Spleen Tyrosine Kinase (SYK), and Factor Xa (FXa). A particularly strong feature of the GSK data is its large size, which lends greater statistical significance to comparisons between different methods. In Phase 1 of the CSAR 2014 Exercise, participants were given several protein-ligand complexes and asked to identify the one near-native pose from among 200 decoys provided by CSAR. Though decoys were requested by the community, we found that they complicated our analysis. We could not discern whether poor predictions were failures of the chosen method or an incompatibility between the participant's method and the setup protocol we used. This problem is inherent to decoys, and we strongly advise against their use. In Phase 2, participants had to dock and rank/score a set of small molecules given only the SMILES strings of the ligands and a protein structure with a different ligand bound. Overall, docking was a success for most participants, much better in Phase 2 than in Phase 1. However, scoring was a greater challenge. No particular approach to docking and scoring had an edge, and successful methods included empirical, knowledge-based, machine-learning, shape-fitting, and even those with solvation and entropy terms. Several groups were successful in ranking TrmD and/or SYK, but ranking FXa ligands was intractable for all participants. Methods that were able to dock well across all submitted systems include MDock,1 Glide-XP,2 PLANTS,3 Wilma,4 Gold,5 SMINA,6 Glide-XP2/PELE,7 FlexX,8 and MedusaDock.9 In fact, the submission based on Glide-XP2/PELE7 cross-docked all ligands to many crystal structures, and it was particularly impressive to see success across an ensemble of protein structures for multiple targets. For scoring/ranking, submissions that showed statistically significant achievement include MDock1 using ITScore1,10 with a flexible-ligand term,11 SMINA6 using Autodock-Vina,12,13 FlexX8 using HYDE,14 and Glide-XP2 using XP DockScore2 with and without ROCS15 shape similarity.16 Of course, these results are for only three protein targets, and many more systems need to be investigated to truly identify which approaches are more successful than others. Furthermore, our exercise is not a competition.

CSAR Benchmark Exercise 2013: Evaluation of Results from a Combined Computational Protein Design, Docking, and Scoring/Ranking Challenge

Community Structure-Activity Resource (CSAR) conducted a benchmark exercise to evaluate the current computational methods for protein design, ligand docking, and scoring/ranking. The exercise consisted of three phases. The first phase required the participants to identify and rank order which designed sequences were able to bind the small molecule digoxigenin. The second phase challenged the community to select a near-native pose of digoxigenin from a set of decoy poses for two of the designed proteins. The third phase investigated the ability of current methods to rank/score the binding affinity of 10 related steroids to one of the designed proteins (pKd = 4.1 to 6.7). We found that 11 of 13 groups were able to correctly select the sequence that bound digoxigenin, with most groups providing the correct three-dimensional structure for the backbone of the protein as well as all atoms of the active-site residues. Eleven of the 14 groups were able to select the appropriate pose from a set of plausible decoy poses. The ability to predict absolute binding affinities is still a difficult task, as 8 of 14 groups were able to correlate scores to affinity (Pearson-r > 0.7) of the designed protein for congeneric steroids and only 5 of 14 groups were able to correlate the ranks of the 10 related ligands (Spearman-ρ > 0.7).

Improvements, trends, and new ideas in molecular docking: 2012-2013 in review

Molecular docking is a computational method for predicting the placement of ligands in the binding sites of their receptor(s). In this review, we discuss the methodological developments that occurred in the docking field in 2012 and 2013, with a particular focus on the more difficult aspects of this computational discipline. The main challenges and therefore focal points for developments in docking, covered in this review, are receptor flexibility, solvation, scoring, and virtual screening. We specifically deal with such aspects of molecular docking and its applications as selection criteria for constructing receptor ensembles, target dependence of scoring functions, integration of higher-level theory into scoring, implicit and explicit handling of solvation in the binding process, and comparison and evaluation of docking and scoring methods.

Cell-based and virtual fragment screening for adrenergic α2C receptor agonists

Fragment-based drug discovery has emerged as an alternative to conventional lead identification and optimization strategies generally supported by biophysical detection techniques. Membrane targets like G protein-coupled receptors (GPCRs), however, offer challenges in lack of generic immobilization or stabilization methods for the dynamic, membrane-bound supramolecular complexes. Also modeling of different functional states of GPCRs proved to be a challenging task. Here we report a functional cell-based high concentration screening campaign for the identification of adrenergic α2C receptor agonists compared with the virtual screening of the same ligand set against an active-like homology model of the α2C receptor. The conventional calcium mobilization-based assay identified active fragments with a similar incidence to several other reported fragment screens on GPCRs. 16 out of 3071 screened fragments turned out as specific ligands of α2C, two of which were identified by virtual screening as well and several of the hits possessed surprisingly high affinity and ligand efficiency. Our results indicate that in vitro biological assays can be utilized in the fragment hit identification process for GPCR targets.

Small-molecule inhibitors of the pseudaminic acid biosynthetic pathway: targeting motility as a key bacterial virulence factor

Helicobacter pylori is motile by means of polar flagella, and this motility has been shown to play a critical role in pathogenicity. The major structural flagellin proteins have been shown to be glycosylated with the nonulosonate sugar, pseudaminic acid (Pse). This glycan is unique to microorganisms, and the process of flagellin glycosylation is required for H. pylori flagellar assembly and consequent motility. As such, the Pse biosynthetic pathway offers considerable potential as an antivirulence drug target, especially since motility is required for H. pylori colonization and persistence in the host. This report describes screening the five Pse biosynthetic enzymes for small-molecule inhibitors using both high-throughput screening (HTS) and in silico (virtual screening [VS]) approaches. Using a 100,000-compound library, 1,773 hits that exhibited a 40% threshold inhibition at a 10 μM concentration were identified by HTS. In addition, VS efforts using a 1.6-million compound library directed at two pathway enzymes identified 80 hits, 4 of which exhibited reasonable inhibition at a 10 μM concentration in vitro. Further secondary screening which identified 320 unique molecular structures or validated hits was performed. Following kinetic studies and structure-activity relationship (SAR) analysis of selected inhibitors from our refined list of 320 compounds, we demonstrated that three inhibitors with 50% inhibitory concentrations (IC50s) of approximately 14 μM, which belonged to a distinct chemical cluster, were able to penetrate the Gram-negative cell membrane and prevent formation of flagella.

Blind prediction of HIV integrase binding from the SAMPL4 challenge

Here, we give an overview of the protein-ligand binding portion of the Statistical Assessment of Modeling of Proteins and Ligands 4 (SAMPL4) challenge, which focused on predicting binding of HIV integrase inhibitors in the catalytic core domain. The challenge encompassed three components--a small "virtual screening" challenge, a binding mode prediction component, and a small affinity prediction component. Here, we give summary results and statistics concerning the performance of all submissions at each of these challenges. Virtual screening was particularly challenging here in part because, in contrast to more typical virtual screening test sets, the inactive compounds were tested because they were thought to be likely binders, so only the very top predictions performed significantly better than random. Pose prediction was also quite challenging, in part because inhibitors in the set bind to three different sites, so even identifying the correct binding site was challenging. Still, the best methods managed low root mean squared deviation predictions in many cases. Here, we give an overview of results, highlight some features of methods which worked particularly well, and refer the interested reader to papers in this issue which describe specific submissions for additional details.

Exhaustive docking and solvated interaction energy scoring: lessons learned from the SAMPL4 challenge

We continued prospective assessments of the Wilma-solvated interaction energy (SIE) platform for pose prediction, binding affinity prediction, and virtual screening on the challenging SAMPL4 data sets including the HIV-integrase inhibitor and two host-guest systems. New features of the docking algorithm and scoring function are tested here prospectively for the first time. Wilma-SIE provides good correlations with actual binding affinities over a wide range of binding affinities that includes strong binders as in the case of SAMPL4 host-guest systems. Absolute binding affinities are also reproduced with appropriate training of the scoring function on available data sets or from comparative estimation of the change in target's vibrational entropy. Even when binding modes are known, SIE predictions lack correlation with experimental affinities within dynamic ranges below 2 kcal/mol as in the case of HIV-integrase ligands, but they correctly signaled the narrowness of the dynamic range. Using a common protein structure for all ligands can reduce the noise, while incorporating a more sophisticated solvation treatment improves absolute predictions. The HIV-integrase virtual screening data set consists of promiscuous weak binders with relatively high flexibility and thus it falls outside of the applicability domain of the Wilma-SIE docking platform. Despite these difficulties, unbiased docking around three known binding sites of the enzyme resulted in over a third of ligands being docked within 2 Å from their actual poses and over half of the ligands docked in the correct site, leading to better-than-random virtual screening results.

Absolute Free Energy of Binding and Entropy of the FKBP12-FK506 Complex: Effects of the Force Field

The hypothetical scanning molecular dynamics (HSMD) method combined with thermodynamic integration (HSMD-TI) has been extended recently for calculating ΔA(0)-the absolute free energy of binding of a ligand to a protein. With HSMD-TI, ΔA(0) is obtained in a new way as a sum of several components, among them is ΔSligand-the change in the conformational entropy as the ligand is transferred from the bulk solvent to the active site-this entropy is obtained by a specific reconstruction procedure. This unique aspect of HSMD (which is useful in rational drug design) is in particular important for treating large ligands, where ΔSligand might be significant. Technically, one should verify that the results for ΔSligand converge-a property that might become more difficult for large ligands; therefore, studying ligands of increasing size would define the range of applicability of HSMD-TI for binding. In this paper, we check the performance of HSMD-TI by applying it to the relatively large ligand FK506 (126 atoms) complexed with the protein FKBP12, where ΔA(0) = -12.8 kcal/mol is known experimentally as well as the crystal structure of the complex. This structure was initially equilibrated by carrying out a 100 ns molecular dynamics trajectory, where the system is modeled by the AMBER force field, TIP3P water, and Particle Mesh Ewald. HSMD-TI calculations were carried out in three conformational regions defined by the intervals [0.2,2], [2,5], and [5,100] ns along the trajectory, where local equilibration of the total energy has been observed; we obtained ΔA(0) = -13.6 ± 1.1, -16.6 ± 1.4, and -16.7 ± 1.4 kcal/mol, respectively indicating the following: (1) The second and third regions belong to the same conformational subspace of the complex, which is different from the [0.2,2] ns subspace. (2) The unsatisfactory result for ΔA(0) obtained in the well equilibrated (hence theoretically preferred) latter regions reflects the nonperfect modeling used, which however (3) has led to the experimental ΔA(0) in the [0.2,2] ns region close to the crystal structure. Keeping the complex near its crystal structure has been a successful approach in the literature. To check this avenue further, we applied harmonic restraints on backbone atoms and obtained unsatisfactory results for ΔA(0), suggesting that implementation of this approach is not straightforward. Converging results for ΔSligand were obtained in all regions, where the result ΔSligand([0.2,2]) = 7.1 ± 1.2 kcal/mol is less region dependent than ΔA(0) and is relatively large probably due to the large ligand.