Enabling structure-based drug discovery utilizing predicted models.

High-quality predicted structures enable structure-based approaches to an expanding number of drug discovery programs. We propose that by utilizing free energy perturbation (FEP), predicted structures can be confidently employed to achieve drug design goals. We use structure-based modeling of hERG inhibition to illustrate this value of FEP.

Free Energy Perturbation Approach for Accurate Crystalline Aqueous Solubility Predictions.

Early assessment of crystalline thermodynamic solubility continues to be elusive for drug discovery and development despite its critical importance, especially for the ever-increasing fraction of poorly soluble drug candidates. Here we present a detailed evaluation of a physics-based free energy perturbation (FEP+) approach for computing the thermodynamic aqueous solubility. The predictive power of this approach is assessed across diverse chemical spaces, spanning pharmaceutically relevant literature compounds and more complex AbbVie compounds. Our approach achieves predictive (RMSE = 0.86) and differentiating power (R2 = 0.69) and therefore provides notably improved correlations to experimental solubility compared to state-of-the-art machine learning approaches that utilize quantum mechanics-based descriptors. The importance of explicit considerations of crystalline packing in predicting solubility by the FEP+ approach is also highlighted in this study. Finally, we show how computed energetics, including hydration and sublimation free energies, can provide further insights into molecule design to feed the medicinal chemistry DMTA cycle.

The maximal and current accuracy of rigorous protein-ligand binding free energy calculations.

Computational techniques can speed up the identification of hits and accelerate the development of candidate molecules for drug discovery. Among techniques for predicting relative binding affinities, the most consistently accurate is free energy perturbation (FEP), a class of rigorous physics-based methods. However, uncertainty remains about how accurate FEP is and can ever be. Here, we present what we believe to be the largest publicly available dataset of proteins and congeneric series of small molecules, and assess the accuracy of the leading FEP workflow. To ascertain the limit of achievable accuracy, we also survey the reproducibility of experimental relative affinity measurements. We find a wide variability in experimental accuracy and a correspondence between binding and functional assays. When careful preparation of protein and ligand structures is undertaken, FEP can achieve accuracy comparable to experimental reproducibility. Throughout, we highlight reliable protocols that can help maximize the accuracy of FEP in prospective studies.

FEP Protocol Builder: Optimization of Free Energy Perturbation Protocols Using Active Learning.

Significant improvements have been made in the past decade to methods that rapidly and accurately predict binding affinity through free energy perturbation (FEP) calculations. This has been driven by recent advances in small-molecule force fields and sampling algorithms combined with the availability of low-cost parallel computing. Predictive accuracies of ∼1 kcal mol-1 have been regularly achieved, which are sufficient to drive potency optimization in modern drug discovery campaigns. Despite the robustness of these FEP approaches across multiple target classes, there are invariably target systems that do not display expected performance with default FEP settings. Traditionally, these systems required labor-intensive manual protocol development to arrive at parameter settings that produce a predictive FEP model. Due to the (a) relatively large parameter space to be explored, (b) significant compute requirements, and (c) limited understanding of how combinations of parameters can affect FEP performance, manual FEP protocol optimization can take weeks to months to complete, and often does not involve rigorous train-test set splits, resulting in potential overfitting. These manual FEP protocol development timelines do not coincide with tight drug discovery project timelines, essentially preventing the use of FEP calculations for these target systems. Here, we describe an automated workflow termed FEP Protocol Builder (FEP-PB) to rapidly generate accurate FEP protocols for systems that do not perform well with default settings. FEP-PB uses an active-learning workflow to iteratively search the protocol parameter space to develop accurate FEP protocols. To validate this approach, we applied it to pharmaceutically relevant systems where default FEP settings could not produce predictive models. We demonstrate that FEP-PB can rapidly generate accurate FEP protocols for the previously challenging MCL1 system with limited human intervention. We also apply FEP-PB in a real-world drug discovery setting to generate an accurate FEP protocol for the p97 system. FEP-PB is able to generate a more accurate protocol than the expert user, rapidly validating p97 as amenable to free energy calculations. Additionally, through the active-learning workflow, we are able to gain insight into which parameters are most important for a given system. These results suggest that FEP-PB is a robust tool that can aid in rapidly developing accurate FEP protocols and increasing the number of targets that are amenable to the technology.

Discovery of a Potent and Selective Tyrosine Kinase 2 Inhibitor: TAK-279.

TYK2 is a key mediator of IL12, IL23, and type I interferon signaling, and these cytokines have been implicated in the pathogenesis of multiple inflammatory and autoimmune diseases such as psoriasis, rheumatoid arthritis, lupus, and inflammatory bowel diseases. Supported by compelling data from human genome-wide association studies and clinical results, TYK2 inhibition through small molecules is an attractive therapeutic strategy to treat these diseases. Herein, we report the discovery of a series of highly selective pseudokinase (Janus homology 2, JH2) domain inhibitors of TYK2 enzymatic activity. A computationally enabled design strategy, including the use of FEP+, was instrumental in identifying a pyrazolo-pyrimidine core. We highlight the utility of computational physics-based predictions used to optimize this series of molecules to identify the development candidate 30, a potent, exquisitely selective cellular TYK2 inhibitor that is currently in Phase 2 clinical trials for the treatment of psoriasis and psoriatic arthritis.

Enhancing Hit Discovery in Virtual Screening through Absolute Protein-Ligand Binding Free-Energy Calculations.

In the hit identification stage of drug discovery, a diverse chemical space needs to be explored to identify initial hits. Contrary to empirical scoring functions, absolute protein-ligand binding free-energy perturbation (ABFEP) provides a theoretically more rigorous and accurate description of protein-ligand binding thermodynamics and could, in principle, greatly improve the hit rates in virtual screening. In this work, we describe an implementation of an accurate and reliable ABFEP method in FEP+. We validated the ABFEP method on eight congeneric compound series binding to eight protein receptors including both neutral and charged ligands. For ligands with net charges, the alchemical ion approach is adopted to avoid artifacts in electrostatic potential energy calculations. The calculated binding free energies correlate with experimental results with a weighted average of R2 = 0.55 for the entire dataset. We also observe an overall root-mean-square error (RMSE) of 1.1 kcal/mol after shifting the zero-point of the simulation data to match the average experimental values. Through ABFEP calculations using apo versus holo protein structures, we demonstrated that the protein conformational and protonation state changes between the apo and holo proteins are the main physical factors contributing to the protein reorganization free energy manifested by the overestimation of raw ABFEP calculated binding free energies using the holo structures of the proteins. Furthermore, we performed ABFEP calculations in three virtual screening applications for hit enrichment. ABFEP greatly improves the hit rates as compared to docking scores or other methods like metadynamics. The good performance of ABFEP in rank ordering compounds demonstrated in this work confirms it as a useful tool to improve the hit rates in virtual screening, thus facilitating hit discovery.

Quantitatively Accounting for Protein Reorganization in Computer-Aided Drug Design.

Structure-based drug design frequently operates under the assumption that a single holo structure is relevant. However, a large number of crystallographic examples clearly show that multiple conformations are possible. In those cases, the protein reorganization free energy must be known to accurately predict binding free energies for ligands. Only then can the energetic preference among these multiple protein conformations be utilized to design ligands with stronger binding potency and selectivity. Here, we present a computational method to quantify these protein reorganization free energies. We test it on two retrospective drug design cases, Abl kinase and HSP90, and illustrate how alternative holo conformations can be derisked and lead to large boosts in affinity. This method will allow computer-aided drug design to better support complex protein targets.

High-Dimensional Neural Network Potential for Liquid Electrolyte Simulations.

Liquid electrolytes are one of the most important components of Li-ion batteries, which are a critical technology of the modern world. However, we still lack the computational tools required to accurately calculate key properties of these materials (viscosity and ionic diffusivity) from first principles necessary to support improved designs. In this work, we report a machine learning-based force field for liquid electrolyte simulations, which bridges the gap between the accuracy of range-separated hybrid density functional theory and the efficiency of classical force fields. Predictions of material properties made with this force field are quantitatively accurate compared to experimental data. Our model uses the QRNN deep neural network architecture, which includes both long-range interactions and global charge equilibration. The training data set is composed solely of non-periodic density functional theory (DFT), allowing the practical use of an accurate theory (here, ωB97X-D3BJ/def2-TZVPD), which would be prohibitively expensive for generating large data sets with periodic DFT. In this report, we focus on seven common carbonates and LiPF6, but this methodology has very few assumptions and can be readily applied to any liquid electrolyte system. This provides a promising path forward for large-scale atomistic modeling of many important battery chemistries.

Induced-Fit Docking Enables Accurate Free Energy Perturbation Calculations in Homology Models.

Homology models have been used for virtual screening and to understand the binding mode of a known active compound; however, rarely have the models been shown to be of sufficient accuracy, comparable to crystal structures, to support free-energy perturbation (FEP) calculations. We demonstrate here that the use of an advanced induced-fit docking methodology reliably enables predictive FEP calculations on congeneric series across homology models ≥30% sequence identity. Furthermore, we show that retrospective FEP calculations on a congeneric series of drug-like ligands are sufficient to discriminate between predicted binding modes. Results are presented for a total of 29 homology models for 14 protein targets, showing FEP results comparable to those obtained using experimentally determined crystal structures for 86% of homology models with template structure sequence identities ranging from 30 to 50%. Implications for the use and validation of homology models in drug discovery projects are discussed.

Exploring the Activity Profile of TbrPDEB1 and hPDE4 Inhibitors Using Free Energy Perturbation.

Human African trypanosomiasis (HAT) is a neglected tropical disease caused by the parasite Trypanosoma brucei (T.b.). A validated target for the treatment of HAT is the parasitic T.b. cyclic nucleotide phosphodiesterase B1 (TbrPDEB1). Although nanomolar TbrPDEB1 inhibitors have been obtained, their activity against the off-target human PDE4 (hPDE4) is likely to lead to undesirable clinical side effects, such as nausea, emesis, and immune suppression. Thus, new and more selective TbrPDEB1 inhibitors are still needed. This retrospective study evaluated the free energy perturbation (FEP+) method to predict the affinity profiles of TbrPDEB1 inhibitors against hPDE4. We demonstrate that FEP+ can be used to accurately predict the activity profiles of these homologous proteins. Moreover, we show how FEP+ can overcome challenges like protein flexibility and high sequence conservation. This also implies that the method can be applied prospectively for the lead optimization campaigns to design new and more selective TbrPDEB1 inhibitors.

AutoDesigner, a De Novo Design Algorithm for Rapidly Exploring Large Chemical Space for Lead Optimization: Application to the Design and Synthesis of d-Amino Acid Oxidase Inhibitors.

The lead optimization stage of a drug discovery program generally involves the design, synthesis, and assaying of hundreds to thousands of compounds. The design phase is usually carried out via traditional medicinal chemistry approaches and/or structure-based drug design (SBDD) when suitable structural information is available. Two of the major limitations of this approach are (1) difficulty in rapidly designing potent molecules that adhere to myriad project criteria, or the multiparameter optimization (MPO) problem, and (2) the relatively small number of molecules explored compared to the vast size of chemical space. To address these limitations, we have developed AutoDesigner, a de novo design algorithm. AutoDesigner employs a cloud-native, multistage search algorithm to carry out successive rounds of chemical space exploration and filtering. Millions to billions of virtual molecules are explored and optimized while adhering to a customizable set of project criteria such as physicochemical properties and potency. Additionally, the algorithm only requires a single ligand with measurable affinity and a putative binding model as a starting point, making it amenable to the early stages of an SBDD project where limited data are available. To assess the effectiveness of AutoDesigner, we applied it to the design of novel inhibitors of d-amino acid oxidase (DAO), a target for the treatment of schizophrenia. AutoDesigner was able to generate and efficiently explore over 1 billion molecules to successfully address a variety of project goals. The compounds generated by AutoDesigner that were synthesized and assayed (1) simultaneously met not only physicochemical criteria, clearance, and central nervous system (CNS) penetration (Kp,uu) cutoffs but also potency thresholds and (2) fully utilize structural data to discover and explore novel interactions and a previously unexplored subpocket in the DAO active site. The reported data demonstrate that AutoDesigner can play a key role in accelerating the discovery of novel, potent chemical matter within the constraints of a given drug discovery lead optimization campaign.

Transferable Neural Network Potential Energy Surfaces for Closed-Shell Organic Molecules: Extension to Ions.

Transferable high dimensional neural network potentials (HDNNPs) have shown great promise as an avenue to increase the accuracy and domain of applicability of existing atomistic force fields for organic systems relevant to life science. We have previously reported such a potential (Schrödinger-ANI) that has broad coverage of druglike molecules. We extend that work here to cover ionic and zwitterionic druglike molecules expected to be relevant to drug discovery research activities. We report a novel HDNNP architecture, which we call QRNN, that predicts atomic charges and uses these charges as descriptors in an energy model that delivers conformational energies within chemical accuracy when measured against the reference theory it is trained to. Further, we find that delta learning based on a semiempirical level of theory approximately halves the errors. We test the models on torsion energy profiles, relative conformational energies, geometric parameters, and relative tautomer errors.

The Impact of Experimental and Calculated Error on the Performance of Affinity Predictions.

The accurate prediction of binding affinity between protein and small molecules with free energy methods, particularly the difference in binding affinities via relative binding free energy calculations, has undergone a dramatic increase in use and impact over recent years. The improvements in methodology, hardware, and implementation can deliver results with less than 1 kcal/mol mean unsigned error between calculation and experiment. This is a remarkable achievement and beckons some reflection on the significance of calculation approaching the accuracy of experiment. In this article, we describe a statistical analysis of the implications of variance (standard deviation) of both experimental and calculated binding affinities with respect to the unknown true binding affinity. We reveal that plausible ratios of standard deviation in experiment and calculation can lead to unexpected outcomes for assessing the performance of predictions. The work extends beyond the case of binding free energies to other affinity or property prediction methods.

From computer-aided drug discovery to computer-driven drug discovery.

Computational chemistry and structure-based design have traditionally been viewed as a subset of tools that could aid acceleration of the drug discovery process, but were not commonly regarded as a driving force in small molecule drug discovery. In the last decade however, there have been dramatic advances in the field, including (1) development of physics-based computational approaches to accurately predict a broad variety of endpoints from potency to solubility, (2) improvements in artificial intelligence and deep learning methods and (3) dramatic increases in computational power with the advent of GPUs and cloud computing, resulting in the ability to explore and accurately profile vast amounts of drug-like chemical space in silico. There have also been simultaneous advancements in structural biology such as cryogenic electron microscopy (cryo-EM) and computational protein-structure prediction, allowing for access to many more high-resolution 3D structures of novel drug-receptor complexes. The convergence of these breakthroughs has positioned structurally-enabled computational methods to be a driving force behind the discovery of novel small molecule therapeutics. This review will give a broad overview of the synergies in recent advances in the fields of computational chemistry, machine learning and structural biology, in particular in the areas of hit identification, hit-to-lead, and lead optimization.

Efficient Exploration of Chemical Space with Docking and Deep Learning.

With the advent of make-on-demand commercial libraries, the number of purchasable compounds available for virtual screening and assay has grown explosively in recent years, with several libraries eclipsing one billion compounds. Today's screening libraries are larger and more diverse, enabling the discovery of more-potent hit compounds and unlocking new areas of chemical space, represented by new core scaffolds. Applying physics-based in silico screening methods in an exhaustive manner, where every molecule in the library must be enumerated and evaluated independently, is increasingly cost-prohibitive. Here, we introduce a protocol for machine learning-enhanced molecular docking based on active learning to dramatically increase throughput over traditional docking. We leverage a novel selection protocol that strikes a balance between two objectives: (1) identifying the best scoring compounds and (2) exploring a large region of chemical space, demonstrating superior performance compared to a purely greedy approach. Together with automated redocking of the top compounds, this method captures almost all the high scoring scaffolds in the library found by exhaustive docking. This protocol is applied to our recent virtual screening campaigns against the D4 and AMPC targets that produced dozens of highly potent, novel inhibitors, and a blind test against the MT1 target. Our protocol recovers more than 80% of the experimentally confirmed hits with a 14-fold reduction in compute cost, and more than 90% of the hit scaffolds in the top 5% of model predictions, preserving the diversity of the experimentally confirmed hit compounds.

OPLS4: Improving Force Field Accuracy on Challenging Regimes of Chemical Space.

We report on the development and validation of the OPLS4 force field. OPLS4 builds upon our previous work with OPLS3e to improve model accuracy on challenging regimes of drug-like chemical space that includes molecular ions and sulfur-containing moieties. A novel parametrization strategy for charged species, which can be extended to other systems, is introduced. OPLS4 leads to improved accuracy on benchmarks that assess small-molecule solvation and protein-ligand binding.

Reliable and Accurate Solution to the Induced Fit Docking Problem for Protein-Ligand Binding.

We present a reliable and accurate solution to the induced fit docking problem for protein-ligand binding by combining ligand-based pharmacophore docking, rigid receptor docking, and protein structure prediction with explicit solvent molecular dynamics simulations. This novel methodology in detailed retrospective and prospective testing succeeded to determine protein-ligand binding modes with a root-mean-square deviation within 2.5 Å in over 90% of cross-docking cases. We further demonstrate these predicted ligand-receptor structures were sufficiently accurate to prospectively enable predictive structure-based drug discovery for challenging targets, substantially expanding the domain of applicability for such methods.

General Theory of Fragment Linking in Molecular Design: Why Fragment Linking Rarely Succeeds and How to Improve Outcomes.

Linking two fragments binding in nearby subpockets together has become an important technique in fragment-based drug discovery to optimize the binding potency of fragment hits. Despite the expected favorable translational and orientational entropic contribution to the binding free energy of the linked molecule, brute force enumeration of chemical linker for linking fragments is rarely successful, and the vast majority of linked molecules do not exhibit the expected gains of binding potency. In this paper, we examine the physical factors that contribute to the change of binding free energy from fragment linking and develop a method to rigorously calculate these different physical contributions. We find from these analyses that multiple confounding factors make successful fragment linking strategies rare, including (1) possible change of the binding mode of the fragments in the linked state compared to separate binding of the fragments, (2) unfavorable intramolecular strain energy of the bioactive conformation of the linked molecule, (3) unfavorable interaction between the linker and the protein, (4) favorable interaction energies between two fragments in solution when not chemically linked that offset the expected entropy loss for the formation of fragment pair, (5) complex compensating configurational entropic effects beyond the simplistic rotational and translational analysis. We here have applied a statistically mechanically rigorous approach to compute the fragment linking coefficients of 10 pharmaceutically interesting systems and quantify the contribution of each physical component to the binding free energy of the linked molecule. Based on these studies, we have found that the change in the relative configurational entropy of the two fragments in the protein binding pocket (a term neglected to our knowledge in all previous analyses) substantially offsets the favorable expected rotational and translational entropic contributions to the binding free energy of the linked molecule. This configurational restriction of the fragments in the binding pocket of the proteins is found to be, in our analysis, the dominant reason why most fragment linking strategies do not exhibit the expected gains of binding potency. These findings have further provided rich physical insights, which we expect should facilitate more successful fragment linking strategies to be formulated in the future.

Is Structure-Based Drug Design Ready for Selectivity Optimization?

Alchemical free-energy calculations are now widely used to drive or maintain potency in small-molecule lead optimization with a roughly 1 kcal/mol accuracy. Despite this, the potential to use free-energy calculations to drive optimization of compound selectivity among two similar targets has been relatively unexplored in published studies. In the most optimistic scenario, the similarity of binding sites might lead to a fortuitous cancellation of errors and allow selectivity to be predicted more accurately than affinity. Here, we assess the accuracy with which selectivity can be predicted in the context of small-molecule kinase inhibitors, considering the very similar binding sites of human kinases CDK2 and CDK9 as well as another series of ligands attempting to achieve selectivity between the more distantly related kinases CDK2 and ERK2. Using a Bayesian analysis approach, we separate systematic from statistical errors and quantify the correlation in systematic errors between selectivity targets. We find that, in the CDK2/CDK9 case, a high correlation in systematic errors suggests that free-energy calculations can have significant impact in aiding chemists in achieving selectivity, while in more distantly related kinases (CDK2/ERK2), the correlation in systematic error suggests that fortuitous cancellation may even occur between systems that are not as closely related. In both cases, the correlation in systematic error suggests that longer simulations are beneficial to properly balance statistical error with systematic error to take full advantage of the increase in apparent free-energy calculation accuracy in selectivity prediction.

Enhancing Water Sampling in Free Energy Calculations with Grand Canonical Monte Carlo.

The prediction of protein-ligand binding affinities using free energy perturbation (FEP) is becoming increasingly routine in structure-based drug discovery. Most FEP packages use molecular dynamics (MD) to sample the configurations of proteins and ligands, as MD is well-suited to capturing coupled motion. However, MD can be prohibitively inefficient at sampling water molecules that are buried within binding sites, which has severely limited the domain of applicability of FEP and its prospective usage in drug discovery. In this paper, we present an advancement of FEP that augments MD with grand canonical Monte Carlo (GCMC), an enhanced sampling method, to overcome the problem of sampling water. We accomplished this without degrading computational performance. On both old and newly assembled data sets of protein-ligand complexes, we show that the use of GCMC in FEP is essential for accurate and robust predictions for ligand perturbations that disrupt buried water.