Quantum mechanical pairwise decomposition analysis of protein kinase B inhibitors: validating a new tool for guiding drug design

The critical role of QM/MM X-ray refinement and accurate tautomer/protomer determination in structure-based drug design

Conventional protein:ligand crystallographic refinement uses stereochemistry restraints coupled with a rudimentary energy functional to ensure the correct geometry of the model of the macromolecule—along with any bound ligand(s)—within the context of the experimental, X-ray density. These methods generally lack explicit terms for electrostatics, polarization, dispersion, hydrogen bonds, and other key interactions, and instead they use pre-determined parameters (e.g. bond lengths, angles, and torsions) to drive structural refinement. In order to address this deficiency and obtain a more complete and ultimately more accurate structure, we have developed an automated approach for macromolecular refinement based on a two layer, QM/MM (ONIOM) scheme as implemented within our DivCon Discovery Suite and "plugged in" to two mainstream crystallographic packages: PHENIX and BUSTER. This implementation is able to use one or more region layer(s), which is(are) characterized using linear-scaling, semi-empirical quantum mechanics, followed by a system layer which includes the balance of the model and which is described using a molecular mechanics functional. In this work, we applied our Phenix/DivCon refinement method—coupled with our XModeScore method for experimental tautomer/protomer state determination—to the characterization of structure sets relevant to structure-based drug design (SBDD). We then use these newly refined structures to show the impact of QM/MM X-ray refined structure on our understanding of function by exploring the influence of these improved structures on protein:ligand binding affinity prediction (and we likewise show how we use post-refinement scoring outliers to inform subsequent X-ray crystallographic efforts). Through this endeavor, we demonstrate a computational chemistry ↔ structural biology (X-ray crystallography) "feedback loop" which has utility in industrial and academic pharmaceutical research as well as other allied fields.

MovableType Software for Fast Free Energy-Based Virtual Screening: Protocol Development, Deployment, Validation, and Assessment

For decades, the complicated energy surfaces found in macromolecular protein:ligand structures, which require large amounts of computational time and resources for energy state sampling, have been an inherent obstacle to fast, routine free energy estimation in industrial drug discovery efforts. Beginning in 2013, the Merz research group addressed this cost with the introduction of a novel sampling methodology termed “Movable Type” (MT). Using numerical integration methods, the MT method reduces the computational expense for energy state sampling by independently calculating each atomic partition function from an initial molecular conformation in order to estimate the molecular free energy using ensembles of the atomic partition functions. In this work, we report a software package, the DivCon Discovery Suite with the MovableType module from QuantumBio Inc., that performs this MT free energy estimation protocol in a fast, fully encapsulated manner. We discuss the computational procedures and improvements to the original work, and we detail the corresponding settings for this software package. Finally, we introduce two validation benchmarks to evaluate the overall robustness of the method against a broad range of protein:ligand structural cases. With these publicly available benchmarks, we show that the method can use a variety of input types and parameters and exhibits comparable predictability whether the method is presented with “expensive” X-ray structures or “inexpensively docked” theoretical models. We also explore some next steps for the method. The MovableType software is available at http://www.quantumbioinc.com/

Insights into the EGFR SAR of N-phenylquinazolin-4-amine-derivatives using quantum mechanical pairwise-interaction energies

Protein kinases are an important class of enzymes that play an essential role in virtually all major disease areas. In addition, they account for approximately 50% of the current targets pursued in drug discovery research. In this work, we explore the generation of structure-based quantum mechanical (QM) quantitative structure-activity relationship models (QSAR) as a means to facilitate structure-guided optimization of protein kinase inhibitors. We explore whether more accurate, interpretable QSAR models can be generated for a series of 76 N-phenylquinazolin-4-amine inhibitors of epidermal growth factor receptor (EGFR) kinase by comparing and contrasting them to other standard QSAR methodologies. The QM-based method involved molecular docking of inhibitors followed by their QM optimization within a ~ 300 atom cluster model of the EGFR active site at the M062X/6-31G(d,p) level. Pairwise computations of the interaction energies with each active site residue were performed. QSAR models were generated by splitting the datasets 75:25 into a training and test set followed by modelling using partial least squares (PLS). Additional QSAR models were generated using alignment dependent CoMFA and CoMSIA methods as well as alignment independent physicochemical, e-state indices and fingerprint descriptors. The structure-based QM-QSAR model displayed good performance on the training and test sets (r² ~ 0.7) and was demonstrably more predictive than the QSAR models built using other methods. The descriptor coefficients from the QM-QSAR models allowed for a detailed rationalization of the active site SAR, which has implications for subsequent design iterations.

High-throughput quantum-mechanics/molecular-mechanics (ONIOM) macromolecular crystallographic refinement with PHENIX/DivCon: the impact of mixed Hamiltonian methods on ligand and protein structure

Conventional macromolecular crystallographic refinement relies on often dubious stereochemical restraints, the preparation of which often requires human validation for unusual species, and on rudimentary energy functionals that are devoid of nonbonding effects owing to electrostatics, polarization, charge transfer or even hydrogen bonding. While this approach has served the crystallographic community for decades, as structure-based drug design/discovery (SBDD) has grown in prominence it has become clear that these conventional methods are less rigorous than they need to be in order to produce properly predictive protein-ligand models, and that the human intervention that is required to successfully treat ligands and other unusual chemistries found in SBDD often precludes high-throughput, automated refinement. Recently, plugins to the Python-based Hierarchical ENvironment for Integrated Xtallography (PHENIX) crystallographic platform have been developed to augment conventional methods with the in situ use of quantum mechanics (QM) applied to ligand(s) along with the surrounding active site(s) at each step of refinement [Borbulevych et al. (2014), Acta Cryst D70, 1233-1247]. This method (Region-QM) significantly increases the accuracy of the X-ray refinement process, and this approach is now used, coupled with experimental density, to accurately determine protonation states, binding modes, ring-flip states, water positions and so on. In the present work, this approach is expanded to include a more rigorous treatment of the entire structure, including the ligand(s), the associated active site(s) and the entire protein, using a fully automated, mixed quantum-mechanics/molecular-mechanics (QM/MM) Hamiltonian recently implemented in the DivCon package. This approach was validated through the automatic treatment of a population of 80 protein-ligand structures chosen from the Astex Diverse Set. Across the entire population, this method results in an average 3.5-fold reduction in ligand strain and a 4.5-fold improvement in MolProbity clashscore, as well as improvements in Ramachandran and rotamer outlier analyses. Overall, these results demonstrate that the use of a structure-wide QM/MM Hamiltonian exhibits improvements in the local structural chemistry of the ligand similar to Region-QM refinement but with significant improvements in the overall structure beyond the active site.

Recent Progress in Treating Protein-Ligand Interactions with Quantum-Mechanical Methods

We review the first successes and failures of a “new wave” of quantum chemistry-based approaches to the treatment of protein/ligand interactions. These approaches share the use of “enhanced”, dispersion (D), and/or hydrogen-bond (H) corrected density functional theory (DFT) or semi-empirical quantum mechanical (SQM) methods, in combination with ensemble weighting techniques of some form to capture entropic effects. Benchmark and model system calculations in comparison to high-level theoretical as well as experimental references have shown that both DFT-D (dispersion-corrected density functional theory) and SQM-DH (dispersion and hydrogen bond-corrected semi-empirical quantum mechanical) perform much more accurately than older DFT and SQM approaches and also standard docking methods. In addition, DFT-D might soon become and SQM-DH already is fast enough to compute a large number of binding modes of comparably large protein/ligand complexes, thus allowing for a more accurate assessment of entropic effects.

XModeScore: a novel method for accurate protonation/tautomer-state determination using quantum-mechanically driven macromolecular X-ray crystallographic refinement

Gaining an understanding of the protein-ligand complex structure along with the proper protonation and explicit solvent effects can be important in obtaining meaningful results in structure-guided drug discovery and structure-based drug discovery. Unfortunately, protonation and tautomerism are difficult to establish with conventional methods because of difficulties in the experimental detection of H atoms owing to the well known limitations of X-ray crystallography. In the present work, it is demonstrated that semiempirical, quantum-mechanics-based macromolecular crystallographic refinement is sensitive to the choice of a protonation-state/tautomer form of ligands and residues, and can therefore be used to explore potential states. A novel scoring method, called XModeScore, is described which enumerates the possible protomeric/tautomeric modes, refines each mode against X-ray diffraction data with the semiempirical quantum-mechanics (PM6) Hamiltonian and scores each mode using a combination of energetic strain (or ligand strain) and rigorous statistical analysis of the difference electron-density distribution. It is shown that using XModeScore it is possible to consistently distinguish the correct bound protomeric/tautomeric modes based on routine X-ray data, even at lower resolutions of around 3 Å. These X-ray results are compared with the results obtained from much more expensive and laborious neutron diffraction studies for three different examples: tautomerism in the acetazolamide ligand of human carbonic anhydrase II (PDB entries 3hs4 and 4k0s), tautomerism in the 8HX ligand of urate oxidase (PDB entries 4n9s and 4n9m) and the protonation states of the catalytic aspartic acid found within the active site of an aspartic protease (PDB entry 2jjj). In each case, XModeScore applied to the X-ray diffraction data is able to determine the correct protonation state as defined by the neutron diffraction data. The impact of QM-based refinement versus conventional refinement on XModeScore is also discussed.

Benchmark of electronic structure methods for protein–ligand interactions based on high-level reference data

The accurate prediction of the strength of protein–ligand interactions is a very difficult problem despite impressive advances in the field of biomolecular modeling. There are good reasons to believe that quantum mechanical methods can help with this task, but the application of such methods in the context of scoring is still in its infancy. Here we benchmark several wave function theory (WFT), density functional theory (DFT) and semiempirical quantum mechanical (SQM) approaches against high-level theoretical references for realistic test cases. Based on our findings for systematically generated model systems of real protein/ligand complexes from the PDB-bind database, we can recommend SCS-MP2 and B2-PLYP-D3 as reference methods, TPSS-D3+Dabc/def-TZVPP as the best DFT approach and PM6-DH+ as a fast and accurate alternative to full ab initio treatments.

Adverse drug reaction prediction using scores produced by large-scale drug-protein target docking on high-performance computing machines

Late-stage or post-market identification of adverse drug reactions (ADRs) is a significant public health issue and a source of major economic liability for drug development. Thus, reliable in silico screening of drug candidates for possible ADRs would be advantageous. In this work, we introduce a computational approach that predicts ADRs by combining the results of molecular docking and leverages known ADR information from DrugBank and SIDER. We employed a recently parallelized version of AutoDock Vina (VinaLC) to dock 906 small molecule drugs to a virtual panel of 409 DrugBank protein targets. L1-regularized logistic regression models were trained on the resulting docking scores of a 560 compound subset from the initial 906 compounds to predict 85 side effects, grouped into 10 ADR phenotype groups. Only 21% (87 out of 409) of the drug-protein binding features involve known targets of the drug subset, providing a significant probe of off-target effects. As a control, associations of this drug subset with the 555 annotated targets of these compounds, as reported in DrugBank, were used as features to train a separate group of models. The Vina off-target models and the DrugBank on-target models yielded comparable median area-under-the-receiver-operating-characteristic-curves (AUCs) during 10-fold cross-validation (0.60-0.69 and 0.61-0.74, respectively). Evidence was found in the PubMed literature to support several putative ADR-protein associations identified by our analysis. Among them, several associations between neoplasm-related ADRs and known tumor suppressor and tumor invasiveness marker proteins were found. A dual role for interstitial collagenase in both neoplasms and aneurysm formation was also identified. These associations all involve off-target proteins and could not have been found using available drug/on-target interaction data. This study illustrates a path forward to comprehensive ADR virtual screening that can potentially scale with increasing number of CPUs to tens of thousands of protein targets and millions of potential drug candidates.

Accurate macromolecular crystallographic refinement: incorporation of the linear scaling, semiempirical quantum-mechanics program DivCon into the PHENIX refinement package

Macromolecular crystallographic refinement relies on sometimes dubious stereochemical restraints and rudimentary energy functionals to ensure the correct geometry of the model of the macromolecule and any covalently bound ligand(s). The ligand stereochemical restraint file (CIF) requires a priori understanding of the ligand geometry within the active site, and creation of the CIF is often an error-prone process owing to the great variety of potential ligand chemistry and structure. Stereochemical restraints have been replaced with more robust functionals through the integration of the linear-scaling, semiempirical quantum-mechanics (SE-QM) program DivCon with the PHENIX X-ray refinement engine. The PHENIX/DivCon package has been thoroughly validated on a population of 50 protein-ligand Protein Data Bank (PDB) structures with a range of resolutions and chemistry. The PDB structures used for the validation were originally refined utilizing various refinement packages and were published within the past five years. PHENIX/DivCon does not utilize CIF(s), link restraints and other parameters for refinement and hence it does not make as many a priori assumptions about the model. Across the entire population, the method results in reasonable ligand geometries and low ligand strains, even when the original refinement exhibited difficulties, indicating that PHENIX/DivCon is applicable to both single-structure and high-throughput crystallography.

Induced fit docking, and the use of QM/MM methods in docking

Docking methods are popular computational techniques in drug discovery to identify new active molecules that bind to a given biological target. Although widely used, the predictive reliability of docking methods is often limited by the inability to accurately and efficiently model protein flexibility and quantify binding strength. We highlight several emerging concepts that address those methodological issues including a discussion on the incorporation of QM/MM methodologies in the scoring process.