Perspective: Alchemical free energy calculations for drug discovery

Adsorption of Flavonoids in a Transcriptional Regulator TtgR: Relative Binding Free Energies and Intermolecular Interactions

Antimicrobial resistance in bacteria often arises from their ability to actively identify and expel toxic compounds. The bacterium strain Pseudomonas putida DOT-T1E utilizes its TtgABC efflux pump to confer robust resistance against antibiotics, flavonoids, and organic solvents. This resistance mechanism is intricately regulated at the transcriptional level by the TtgR protein. Through molecular dynamics and alchemical free energy simulations, we systematically examine the binding of seven flavonoids and their derivatives with the TtgR transcriptional regulator. Our simulations reveal distinct binding geometries and free energies for the flavonoids in the active site of the protein, which are driven by a range of noncovalent forces encompassing van der Waals, electrostatic, and hydrogen bonding interactions. The interplay of molecular structures, substituent patterns, and intermolecular interactions effectively stabilizes the bound flavonoids, confining their movements within the TtgR binding pocket. These findings yield valuable insights into the molecular determinants that govern ligand recognition in TtgR and shed light on the mechanism of antimicrobial resistance in P. putida DOT-T1E.

Current State of Open Source Force Fields in Protein-Ligand Binding Affinity Predictions

In drug discovery, the in silico prediction of binding affinity is one of the major means to prioritize compounds for synthesis. Alchemical relative binding free energy (RBFE) calculations based on molecular dynamics (MD) simulations are nowadays a popular approach for the accurate affinity ranking of compounds. MD simulations rely on empirical force field parameters, which strongly influence the accuracy of the predicted affinities. Here, we evaluate the ability of six different small-molecule force fields to predict experimental protein-ligand binding affinities in RBFE calculations on a set of 598 ligands and 22 protein targets. The public force fields OpenFF Parsley and Sage, GAFF, and CGenFF show comparable accuracy, while OPLS3e is significantly more accurate. However, a consensus approach using Sage, GAFF, and CGenFF leads to accuracy comparable to OPLS3e. While Parsley and Sage are performing comparably based on aggregated statistics across the whole dataset, there are differences in terms of outliers. Analysis of the force field reveals that improved parameters lead to significant improvement in the accuracy of affinity predictions on subsets of the dataset involving those parameters. Lower accuracy can not only be attributed to the force field parameters but is also dependent on input preparation and sampling convergence of the calculations. Especially large perturbations and nonconverged simulations lead to less accurate predictions. The input structures, Gromacs force field files, as well as the analysis Python notebooks are available on GitHub.

DrugGym: A testbed for the economics of autonomous drug discovery

Drug discovery is stochastic. The effectiveness of candidate compounds in satisfying design objectives is unknown ahead of time, and the tools used for prioritization-predictive models and assays-are inaccurate and noisy. In a typical discovery campaign, thousands of compounds may be synthesized and tested before design objectives are achieved, with many others ideated but deprioritized. These challenges are well-documented, but assessing potential remedies has been difficult. We introduce DrugGym, a framework for modeling the stochastic process of drug discovery. Emulating biochemical assays with realistic surrogate models, we simulate the progression from weak hits to sub-micromolar leads with viable ADME. We use this testbed to examine how different ideation, scoring, and decision-making strategies impact statistical measures of utility, such as the probability of program success within predefined budgets and the expected costs to achieve target candidate profile (TCP) goals. We also assess the influence of affinity model inaccuracy, chemical creativity, batch size, and multi-step reasoning. Our findings suggest that reducing affinity model inaccuracy from 2 to 0.5 pIC50 units improves budget-constrained success rates tenfold. DrugGym represents a realistic testbed for machine learning methods applied to the hit-to-lead phase. Source code is available at www.drug-gym.org.

Investigating the structural impact of Omicron RBD mutation on antibody escape and receptor management

The SARS-CoV-2 Variant B.1.1.5291 evolved rapidly in late November 2021 from the existing mutants sparking fear worldwide owing to its infamous immune escape from a varied class of neutralising antibodies. To assess the structural behaviour of Omicron-Receptor Binding Domain (RBD) upon interacting with cross-reactive CR3022 antibody, we investigated the computational approach of structural engagement in B.1.1529 RBD and wild-type RBD with CR3022 antibody. The current study investigates the interacting interface between the RBDs and CR3022 to decipher the key residues accompanying the potential mutational landscape of SARS-CoV-2 variants. We conducted in-silico docking followed by molecular dynamics simulation analysis to examine the dynamic behaviour of protein-protein interactions. Furthermore, the study unleashed possible interactions post energy decomposition analysis via MM-GBSA. Conclusively, the mutational landscape of RBD eases in designing and discovering the effective neutralisation accompanied by development of a universal vaccine.Communicated by Ramaswamy H. Sarma.

Quantitative Prediction of Dissociation Rates of PYK2 Ligands Using Umbrella Sampling and Milestoning

We used umbrella sampling and the milestoning simulation method to study the dissociation of multiple ligands from protein kinase PYK2. The activation barriers obtained from the potential of mean force of the umbrella sampling simulations correlated well with the experimental dissociation rates. Using the zero-temperature string method, we obtained optimized paths along the free-energy surfaces for milestoning simulations of three ligands with a similar structure. The milestoning simulations gave an absolute dissociation rate within 2 orders of magnitude of the experimental value for two ligands but at least 3 orders of magnitude too high for the third. Despite the similarity in their structures, the ligands took different pathways to exit from the binding site of PYK2, making contact with different sets of residues. In addition, the protein experienced different conformational changes for dissociation of the three ligands.

Alchemical Enhanced Sampling with Optimized Phase Space Overlap

An alchemical enhanced sampling (ACES) method has recently been introduced to facilitate importance sampling in free energy simulations. The method achieves enhanced sampling from Hamiltonian replica exchange within a dual topology framework while utilizing new smoothstep softcore potentials. A common sampling problem encountered in lead optimization is the functionalization of aromatic rings that exhibit distinct conformational preferences when interacting with the protein. It is difficult to converge the distribution of ring conformations due to the long time scale of ring flipping events; however, the ACES method addresses this issue by modeling the syn and anti ring conformations within a dual topology. ACES thereby samples the conformer distributions by alchemically tunneling between states, as opposed to traversing a physical pathway with a high rotational barrier. We demonstrate the use of ACES to overcome conformational sampling issues involving ring flipping in ML300-derived noncovalent inhibitors of SARS-CoV-2 Main Protease (Mpro). The demonstrations explore how the use of replica exchange and the choice of softcore selection affects the convergence of the ring conformation distributions. Furthermore, we examine how the accuracy of the calculated free energies is affected by the degree of phase space overlap (PSO) between adjacent states (i.e., between neighboring λ-windows) and the Hamiltonian replica exchange acceptance ratios. Both of these factors are sensitive to the spacing between the intermediate states. We introduce a new method for choosing a schedule of λ values. The method analyzes short "burn-in" simulations to construct a 2D map of the nonlocal PSO. The schedule is obtained by optimizing an alchemical pathway on the 2D map that equalizes the PSO between the λ intervals. The optimized phase space overlap λ-spacing method (Opt-PSO) leads to more numerous end-to-end single passes and round trips due to the correlation between PSO and Hamiltonian replica exchange acceptance ratios. The improved exchange statistics enhance the efficiency of ACES method. The method has been implemented into the FE-ToolKit software package, which is freely available.

Binding Sites of Bicarbonate in Phosphoenolpyruvate Carboxylase

Phosphoenolpyruvate carboxylase (PEPC) is used in plant metabolism for fruit maturation or seed development as well as in the C4 and crassulacean acid metabolism (CAM) mechanisms in photosynthesis, where it is used for the capture of hydrated CO2 (bicarbonate). To find the yet unknown binding site of bicarbonate in this enzyme, we have first identified putative binding sites with nonequilibrium molecular dynamics simulations and then ranked these sites with alchemical free energy calculations with corrections of computational artifacts. Fourteen pockets where bicarbonate could bind were identified, with three having realistic binding free energies with differences with the experimental value below 1 kcal/mol. One of these pockets is found far from the active site at 14 Å and predicted to be an allosteric binding site. In the two other binding sites, bicarbonate is in direct interaction with the magnesium ion; neither sequence alignment nor the study of mutant K606N allowed to discriminate between these two pockets, and both are good candidates as the binding site of bicarbonate in phosphoenolpyruvate carboxylase.

System-Specific Parameter Optimization for Nonpolarizable and Polarizable Force Fields

The accuracy of classical force fields (FFs) has been shown to be limited for the simulation of cation-protein systems despite their importance in understanding the processes of life. Improvements can result from optimizing the parameters of classical FFs or by extending the FF formulation by terms describing charge transfer (CT) and polarization (POL) effects. In this work, we introduce our implementation of the CTPOL model in OpenMM, which extends the classical additive FF formula by adding CT and POL. Furthermore, we present an open-source parametrization tool, called FFAFFURR, that enables the (system-specific) parametrization of OPLS-AA and CTPOL models. The performance of our workflow was evaluated by its ability to reproduce quantum chemistry energies and by molecular dynamics simulations of a zinc-finger protein.

Machine Learning Methods as a Cost-Effective Alternative to Physics-Based Binding Free Energy Calculations

The rank ordering of ligands remains one of the most attractive challenges in drug discovery. While physics-based in silico binding affinity methods dominate the field, they still have problems, which largely revolve around forcefield accuracy and sampling. Recent advances in machine learning have gained traction for protein-ligand binding affinity predictions in early drug discovery programs. In this article, we perform retrospective binding free energy evaluations for 172 compounds from our internal collection spread over four different protein targets and five congeneric ligand series. We compared multiple state-of-the-art free energy methods ranging from physics-based methods with different levels of complexity and conformational sampling to state-of-the-art machine-learning-based methods that were available to us. Overall, we found that physics-based methods behaved particularly well when the ligand perturbations were made in the solvation region, and they did not perform as well when accounting for large conformational changes in protein active sites. On the other end, machine-learning-based methods offer a good cost-effective alternative for binding free energy calculations, but the accuracy of their predictions is highly dependent on the experimental data available for training the model.

Fast free energy estimates from λ-dynamics with bias-updated Gibbs sampling

Relative binding free energy calculations have become an integral computational tool for lead optimization in structure-based drug design. Classical alchemical methods, including free energy perturbation or thermodynamic integration, compute relative free energy differences by transforming one molecule into another. However, these methods have high operational costs due to the need to perform many pairwise perturbations independently. To reduce costs and accelerate molecular design workflows, we present a method called λ-dynamics with bias-updated Gibbs sampling. This method uses dynamic biases to continuously sample between multiple ligand analogues collectively within a single simulation. We show that many relative binding free energies can be determined quickly with this approach without compromising accuracy. For five benchmark systems, agreement to experiment is high, with root mean square errors near or below 1.0 kcal mol-1. Free energy results are consistent with other computational approaches and within statistical noise of both methods (0.4 kcal mol-1 or less). Notably, large efficiency gains over thermodynamic integration of 18-66-fold for small perturbations and 100-200-fold for whole aromatic ring substitutions are observed. The rapid determination of relative binding free energies will enable larger chemical spaces to be more readily explored and structure-based drug design to be accelerated.

Modern Alchemical Free Energy Methods for Drug Discovery Explained

This Perspective provides a contextual explanation of the current state-of-the-art alchemical free energy methods and their role in drug discovery as well as highlights select emerging technologies. The narrative attempts to answer basic questions about what goes on "under the hood" in free energy simulations and provide general guidelines for how to run simulations and analyze the results. It is the hope that this work will provide a valuable introduction to students and scientists in the field.

Can Free Energy Perturbation Simulations Coupled with Replica-Exchange Molecular Dynamics Study Ligands with Distributed Binding Sites?

Free energy perturbation coupled with replica exchange with solute tempering (FEP/REST) offers a rigorous approach to compute relative free energy changes for ligands. To determine the applicability of FEP/REST for the ligands with distributed binding poses, we considered two alchemical transformations involving three putative inhibitors I0, I1, and I2 of the Venezuelan equine encephalitis virus nuclear localization signal sequence binding to the importin-α (impα) transporter protein. I0 → I1 and I0 → I2 transformations, respectively, increase or decrease the polarity of the parent molecule. Our objective was three-fold─(i) to verify FEP/REST technical performance and convergence, (ii) to estimate changes in binding free energy ΔΔG, and (iii) to determine the utility of FEP/REST simulations for conformational binding analysis. Our results are as follows. First, our FEP/REST implementation properly follows FEP/REST formalism and produces converged ΔΔG estimates. Due to ligand inherent unbinding, the better FEP/REST strategy lies in performing multiple independent trajectories rather than extending their length. Second, I0 → I1 and I0 → I2 transformations result in overall minor changes in inhibitor binding free energy, slightly strengthening the affinity of I1 and weakening that of I2. Electrostatic interactions dominate binding interactions, determining the enthalpic changes. The two transformations cause opposite entropic changes, which ultimately govern binding affinities. Importantly, we confirm the validity of FEP/REST free energy estimates by comparing them with our previous REST simulations, directly probing binding of three ligands to impα. Third, we established that FEP/REST simulations can sample binding ensembles of ligands. Thus, FEP/REST can be applied (i) to study the energetics of the ligand binding without defined poses and showing minor differences in affinities |ΔΔG| ≲ 0.5 kcal/mol and (ii) to collect ligand binding conformational ensembles.

Molecular Dynamics Simulations of Matrix Metalloproteinase 13 and the Analysis of the Specificity Loop and the S1'-Site

The specificity loop of Matrix Metalloproteinases (MMPs) is known to regulate recognition of their substrates, and the S1'-site surrounded by the loop is a unique place to address the selectivity of ligands toward each MMP. Molecular dynamics (MD) simulations of apo-MMP-13 and its complex forms with various ligands were conducted to identify the role of the specificity loop for the ligand binding to MMP-13. The MD simulations showed the dual role of T247 as a hydrogen bond donor to the ligand, as well as a contributor to the formation of the van der Waal surface area, with T245 and K249 on the S1'-site. The hydrophobic surface area mediated by T247 blocks the access of water molecules to the S1'-site of MMP-13 and stabilizes the ligand in the site. The F252 residue is flexible in order to search for the optimum location in the S1'-site of the apo-MMP-13, but once a ligand binds to the S1'-site, it can form offset π-π or edge-to-π stacking interactions with the ligand. Lastly, H222 and Y244 provide the offset π-π and π-CH(Cβ) interactions on each side of the phenyl ring of the ligand, and this sandwiched interaction could be critical for the ligand binding to MMP-13.

Accurate Host-Guest Binding Free Energies Using the AMOEBA Polarizable Force Field

A grand challenge of computational biophysics is accurate prediction of interactions between molecules. Molecular dynamics (MD) simulations have recently gained much interest as a tool to directly compute rigorous intermolecular binding affinities. The choice of a fixed point-charge or polarizable multipole force field used in MD is a topic of ongoing discussion. To compare alternative methods, we participated in the SAMPL7 and SAMPL8 Gibb octaacid host-guest challenges to assess the Atomic Multipole Optimized Energetics for Biomolecular Applications (AMOEBA) polarizable multipole force field. Advantages of AMOEBA over fixed charge models include improved representation of molecular electrostatic potentials and better description of water occupying the unligated host cavity. Prospective predictions for 26 host-guest systems exhibit a mean unsigned error vs experiment of 0.848 kcal/mol across all absolute binding free energies, demonstrating excellent agreement between computational and experimental results. In addition, we explore two topics related to the inclusion of ions in MD simulations: use of a neutral co-alchemical protocol and the effect of salt concentration on binding affinity. Use of the co-alchemical method minimally affects computed energies, but salt concentration significantly perturbs our binding results. Higher salt concentration strengthens binding through classical charge screening. In particular, added Na+ ions screen negatively charged carboxylate groups near the binding cavity, thereby diminishing repulsive coulomb interactions with negatively charged guests. Overall, the AMOEBA results demonstrate the accuracy available through a force field providing a detailed energetic description of the four octaacid hosts and 13 charged organic guests. Use of the AMOEBA polarizable atomic multipole force field in conjunction with an alchemical free energy protocol can achieve chemical accuracy in application to realistic molecular systems.

Combining QM/MM Calculations with Classical Mining Minima to Predict Protein-Ligand Binding Free Energy

We developed an effective binding free energy prediction protocol which incorporates quantum mechanical/molecular mechanical (QM/MM) calculations to substitute the specified atomic charges of force fields with quantum-mechanically recalculated ones at a proposed pose using a mining minima approach with the VeraChem mining minima engine. We tested this protocol using seven well-known targets with 147 different ligands and compared it with classical mining minima and the most popular binding free energy (BFE) methods using different metrics. Our new protocol, dubbed Qcharge-VM2, yielded an overall Pearson correlation of 0.86, which was better than all the methods examined. Qcharge-VM2 performed significantly better than implicit solvent-based methods, such as MM-GBSA and MM-PBSA, but not as good as explicit water-based free energy perturbation methods, such as FEP+, in terms of root-mean-square error, RMSE (1.75 kcal/mol) and mean unsigned error, MUE (1.39 kcal/mol) on a limited set of targets. However, our protocol is substantially less computationally demanding compared with FEP+. The combined accuracy and efficiency of our method can be valuable in drug discovery campaigns.

Chemical Promoter Performance for CO2 Hydrate Growth: A Molecular Perspective

Carbon dioxide (CO₂) hydrates, which contain a relatively large amount of captured CO₂ (almost 30 wt % of CO₂ with the balance being water), represent a promising CO₂ sequestration option for climate change mitigation. To facilitate CO₂ storage via hydrates, using chemical additives during hydrate formation might help to expedite formation/growth rates, provided the additives do not reduce the storage capacity. Implementing atomistic molecular dynamics, we study the impact of aziridine, pyrrolidine, and tetrahydrofuran (THF) on the kinetics of CO₂ hydrate growth/dissociation. Our simulations are validated via reproducing experimental data for CO₂ and CO₂ + THF hydrates at selected operating conditions. The simulated results show that both aziridine and pyrrolidine could perform as competent thermodynamic and kinetic promoters. Furthermore, aziridine seems to exceed pyrrolidine and THF in expediting the CO₂ hydrate growth rates under the same conditions. Our analysis unveils direct correlations between the kinetics of CO₂ hydrate growth and a combination of the free energy barrier for desorption of CO₂ from the hydrate surface and the binding free energy of chemical additives adsorbed at the growing hydrate substrate. The detailed thermodynamic analysis conducted in both hydrate and aqueous phases reveals molecular-level mechanisms by which CO₂ hydrate promoters are active, which could help to enable the implementation of CO₂ sequestration in hydrate-bearing reservoirs.

Quality over quantity: Sampling high probability rare events with the weighted ensemble algorithm

The prediction of (un)binding rates and free energies is of great significance to the drug design process. Although many enhanced sampling algorithms and approaches have been developed, there is not yet a reliable workflow to predict these quantities. Previously we have shown that free energies and transition rates can be calculated by directly simulating the binding and unbinding processes with our variant of the WE algorithm "Resampling of Ensembles by Variation Optimization", or "REVO". Here, we calculate binding free energies retrospectively for three SAMPL6 host-guest systems and prospectively for a SAMPL9 system to test a modification of REVO that restricts its cloning behavior in quasi-unbound states. Specifically, trajectories cannot clone if they meet a physical requirement that represents a high likelihood of unbinding, which in the case of this work is a center-of-mass to center-of-mass distance. The overall effect of this change was difficult to predict, as it results in fewer unbinding events each of which with a much higher statistical weight. For all four systems tested, this new strategy produced either more accurate unbinding free energies or more consistent results between simulations than the standard REVO algorithm. This approach is highly flexible, and any feature of interest for a system can be used to determine cloning eligibility. These findings thus constitute an important improvement in the calculation of transition rates and binding free energies with the weighted ensemble method.

To Design Scalable Free Energy Perturbation Networks, Optimal Is Not Enough

Drug discovery is accelerated with computational methods such as alchemical simulations to estimate ligand affinities. In particular, relative binding free energy (RBFE) simulations are beneficial for lead optimization. To use RBFE simulations to compare prospective ligands in silico, researchers first plan the simulation experiment, using graphs where nodes represent ligands and graph edges represent alchemical transformations between ligands. Recent work demonstrated that optimizing the statistical architecture of these perturbation graphs improves the accuracy of the predicted changes in the free energy of ligand binding. Therefore, to improve the success rate of computational drug discovery, we present the open-source software package High Information Mapper (HiMap)─a new take on its predecessor, Lead Optimization Mapper (LOMAP). HiMap removes heuristics decisions from design selection and instead finds statistically optimal graphs over ligands clustered with machine learning. Beyond optimal design generation, we present theoretical insights for designing alchemical perturbation maps. Some of these results include that for n number of nodes, the precision of perturbation maps is stable at n·ln(n) edges. This result indicates that even an "optimal" graph can result in unexpectedly high errors if a plan includes too few alchemical transformations for the given number of ligands and edges. And, as a study compares more ligands, the performance of even optimal graphs will deteriorate with linear scaling of the edge count. In this sense, ensuring an A- or D-optimal topology is not enough to produce robust errors. We additionally find that optimal designs will converge more rapidly than radial and LOMAP designs. Moreover, we derive bounds for how clustering reduces cost for designs with a constant expected relative error per cluster, invariant of the size of the design. These results inform how to best design perturbation maps for computational drug discovery and have broader implications for experimental design.

ACES: Optimized Alchemically Enhanced Sampling

We present an alchemical enhanced sampling (ACES) method implemented in the GPU-accelerated AMBER free energy MD engine. The methods hinges on the creation of an "enhanced sampling state" by reducing or eliminating selected potential energy terms and interactions that lead to kinetic traps and conformational barriers while maintaining those terms that curtail the need to otherwise sample large volumes of phase space. For example, the enhanced sampling state might involve transforming regions of a ligand and/or protein side chain into a noninteracting "dummy state" with internal electrostatics and torsion angle terms turned off. The enhanced sampling state is connected to a real-state end point through a Hamiltonian replica exchange (HREMD) framework that is facilitated by newly developed alchemical transformation pathways and smoothstep softcore potentials. This creates a counterdiffusion of real and enhanced-sampling states along the HREMD network. The effect of a differential response of the environment to the real and enhanced-sampling states is minimized by leveraging the dual topology framework in AMBER to construct a counterbalancing HREMD network in the opposite alchemical direction with the same (or similar) real and enhanced sampling states at inverted end points. The method has been demonstrated in a series of test cases of increasing complexity where traditional MD, and in several cases alternative REST2-like enhanced sampling methods, are shown to fail. The hydration free energy for acetic acid was shown to be independent of the starting conformation, and the values for four additional edge case molecules from the FreeSolv database were shown to have a significantly closer agreement with experiment using ACES. The method was further able to handle different rotamer states in a Cdk2 ligand identified as fractionally occupied in crystal structures. Finally, ACES was applied to T4-lysozyme and demonstrated that the side chain distribution of V111χ1 could be reliably reproduced for the apo state, bound to p-xylene, and in p-xylene→ benzene transformations. In these cases, the ACES method is shown to be highly robust and superior to a REST2-like enhanced sampling implementation alone.

AMBER Free Energy Tools: A New Framework for the Design of Optimized Alchemical Transformation Pathways

We develop a framework for the design of optimized alchemical transformation pathways in free energy simulations using nonlinear mixing and a new functional form for so-called "softcore" potentials. We describe the implementation and testing of this framework in the GPU-accelerated AMBER software suite. The new optimized alchemical transformation pathways integrate a number of important features, including (1) the use of smoothstep functions to stabilize behavior near the transformation end points, (2) consistent power scaling of Coulomb and Lennard-Jones (LJ) interactions with unitless control parameters to maintain balance of electrostatic attractions and exchange repulsions, (3) pairwise form based on the LJ contact radius for the effective interaction distance with separation-shifted scaling, and (4) rigorous smoothing of the potential at the nonbonded cutoff boundary. The new softcore potential form is combined with smoothly transforming nonlinear λ weights for mixing specific potential energy terms, along with flexible λ-scheduling features, to enable robust and stable alchemical transformation pathways. The resulting pathways are demonstrated and tested, and shown to be superior to the traditional methods in terms of numerical stability and minimal variance of the free energy estimates for all cases considered. The framework presented here can be used to design new alchemical enhanced sampling methods, and leveraged in robust free energy workflows for large ligand data sets.

Comprehensive Survey of Consensus Docking for High-Throughput Virtual Screening

The rapid advances of 3D techniques for the structural determination of proteins and the development of numerous computational methods and strategies have led to identifying highly active compounds in computer drug design. Molecular docking is a method widely used in high-throughput virtual screening campaigns to filter potential ligands targeted to proteins. A great variety of docking programs are currently available, which differ in the algorithms and approaches used to predict the binding mode and the affinity of the ligand. All programs heavily rely on scoring functions to accurately predict ligand binding affinity, and despite differences in performance, none of these docking programs is preferable to the others. To overcome this problem, consensus scoring methods improve the outcome of virtual screening by averaging the rank or score of individual molecules obtained from different docking programs. The successful application of consensus docking in high-throughput virtual screening highlights the need to optimize the predictive power of molecular docking methods.

Practical Guidance for Consensus Scoring and Force Field Selection in Protein-Ligand Binding Free Energy Simulations

The advances in ligand binding affinity prediction have been fostered by system generation tools and improved force fields (FFs). CHARMM-GUI Free Energy Calculator provides input and postprocessing scripts for AMBER-TI free energy calculations with various FFs. In this study, we used 12 different FF combinations (ff14SB and ff19SB for protein, GAFF2.2 and OpenFF for ligand, and TIP3P, TIP4PEW, and OPC for water) to calculate relative binding free energies (ΔΔG_bind) for 80 alchemical transformations (among the JACS benchmark set) with different numbers of λ windows (12 or 21) and simulation times (1, 5, or 10 ns). Our results show that 12 λ windows and 5 ns simulation time for each window are sufficient to obtain reliable ΔΔG_bind with 4 independent runs for the current benchmark set. While there is no statistically noticeable performance difference among 12 different FF combinations compared to the experimental values, a combination of ff14SB + GAFF2.2 + TIP3P FFs appears to be best with a mean unsigned error of 0.87 [0.69, 1.07] kcal/mol, a root-mean-square error of 1.22 [0.94, 1.50] kcal/mol, a Pearson's correlation of 0.64 [0.52, 0.76], a Spearman's correlation of 0.73 [0.58, 0.83], and a Kendell's correlation of 0.54 [0.42, 0.64]. This large-scale ΔΔG_bind calculation study provides useful information about ΔΔG_bind prediction with different AMBER FF combinations and presents valuable suggestions for FF selection in AMBER-TI ΔΔG_bind calculations.

AMBER Drug Discovery Boost Tools: Automated Workflow for Production Free-Energy Simulation Setup and Analysis (ProFESSA)

We report an automated workflow for production free-energy simulation setup and analysis (ProFESSA) using the GPU-accelerated AMBER free-energy engine with enhanced sampling features and analysis tools, part of the AMBER Drug Discovery Boost package that has been integrated into the AMBER22 release. The workflow establishes a flexible, end-to-end pipeline for performing alchemical free-energy simulations that brings to bear technologies, including new enhanced sampling features and analysis tools, to practical drug discovery problems. ProFESSA provides the user with top-level control of large sets of free-energy calculations and offers access to the following key functionalities: (1) automated setup of file infrastructure; (2) enhanced conformational and alchemical sampling with the ACES method; and (3) network-wide free-energy analysis with the optional imposition of cycle closure and experimental constraints. The workflow is applied to perform absolute and relative solvation free-energy and relative ligand-protein binding free-energy calculations using different atom-mapping procedures. Results demonstrate that the workflow is internally consistent and highly robust. Further, the application of a new network-wide Lagrange multiplier constraint analysis that imposes key experimental constraints substantially improves binding free-energy predictions.

Absolute Binding Free Energy Calculations for Buried Water Molecules

Water often plays a key role in mediating protein-ligand interactions. Understanding contributions from active-site water molecules to binding thermodynamics of a ligand is important in predicting binding free energies for ligand optimization. In this work, we tested a non-equilibrium switching method for absolute binding free energy calculations on water molecules in binding sites of 13 systems. We discuss the lessons we learned about identified issues that affected our calculations and ways to address them. This work fits with our larger focus on how to do accurate ligand binding free energy calculations when water rearrangements are very slow, such as rearrangements due to ligand modification (as in relative free energy calculations) or ligand binding (as in absolute free energy calculations). The method studied in this work can potentially be used to account for limited water sampling via providing endpoint corrections to free energy calculations using our calculated binding free energy of water.

Simulation of ligand dissociation kinetics from the protein kinase PYK2

Early-stage drug discovery projects often focus on equilibrium binding affinity to the target alongside selectivity and other pharmaceutical properties. The kinetics of drug binding are ignored but can have significant influence on drug efficacy. Therefore, increasing attention has been paid on evaluating drug-binding kinetics early in a drug discovery process. Simulating drug-binding kinetics at the atomic level is challenging for the long time scale involved. Here, we used the transition-based reweighting analysis method (TRAM) with the Markov state model to study the dissociation of a ligand from the protein kinase PYK2. TRAM combines biased and unbiased simulations to reduce computational costs. This work used the umbrella sampling technique for the biased simulations. Although using the potential of mean force from umbrella sampling simulations with the transition-state theory over-estimated the dissociation rate by three orders of magnitude, TRAM gave a dissociation rate within an order of magnitude of the experimental value.

Towards rational computational peptide design

Peptides are prevalent in biology, mediating as many as 40% of protein-protein interactions, and involved in other cellular functions such as transport and signaling. Their ability to bind with high specificity make them promising therapeutical agents with intermediate properties between small molecules and large biologics. Beyond their biological role, peptides can be programmed to self-assembly, and they are already being used for functions as diverse as oligonuclotide delivery, tissue regeneration or as drugs. However, the transient nature of their interactions has limited the number of structures and knowledge of binding affinities available-and their flexible nature has limited the success of computational pipelines that predict the structures and affinities of these molecules. Fortunately, recent advances in experimental and computational pipelines are creating new opportunities for this field. We are starting to see promising predictions of complex structures, thermodynamic and kinetic properties. We believe in the following years this will lead to robust rational peptide design pipelines with success similar to those applied for small molecule drug discovery.

A New Formulation for the Concerted Alchemical Calculation of van der Waals and Coulomb Components of Solvation Free Energies

Alchemical free energy calculations via molecular dynamics have been widely used to obtain thermodynamic properties related to protein-ligand binding and solute-solvent interactions. Although soft-core modeling is the most common approach, the linear basis function (LBF) methodology [Naden, L. N.; et al. J. Chem. Theory Comput.2014, 10 (3), 1128; 2015, 11 (6), 2536] has emerged as a suitable alternative. It overcomes the end-point singularity of the scaling method while maintaining essential advantages such as ease of implementation and high flexibility for postprocessing analysis. In the present work, we propose a simple LBF variant and formulate an efficient protocol for evaluating van der Waals and Coulomb components of an alchemical transformation in tandem, in contrast to the prevalent sequential evaluation mode. To validate our proposal, which results from a careful optimization study, we performed solvation free energy calculations and obtained octanol-water partition coefficients of small organic molecules. Comparisons with results obtained via the sequential mode using either another LBF approach or the soft-core model attest to the effectiveness and correctness of our method. In addition, we show that a reaction field model with an infinite dielectric constant can provide very accurate hydration free energies when used instead of a lattice-sum method to model solute-solvent electrostatics.

Estimating the binding energetics of reversible covalent inhibitors of the SARS-CoV-2 main protease: an in silico study

The main protease (M^pro) of the SARS-CoV-2 virus is an attractive therapeutic target for developing antivirals to combat COVID-19. M^pro is essential for the replication cycle of the SARS-CoV-2 virus, so inhibiting M^pro blocks a vital piece of the cell replication machinery of the virus. A promising strategy to disrupt the viral replication cycle is to design inhibitors that bind to the active site cysteine (Cys145) of the M^pro. Cysteine targeted covalent inhibitors are gaining traction in drug discovery owing to the benefits of improved potency and extended drug-target engagement. An interesting aspect of these inhibitors is that they can be chemically tuned to form a covalent, but reversible bond, with their targets of interest. Several small-molecule cysteine-targeting covalent inhibitors of the M^pro have been discovered-some of which are currently undergoing evaluation in early phase human clinical trials. Understanding the binding energetics of these inhibitors could provide new insights to facilitate the design of potential drug candidates against COVID-19. Motivated by this, we employed rigorous absolute binding free energy calculations and hybrid quantum mechanical/molecular mechanical (QM/MM) calculations to estimate the energetics of binding of some promising reversible covalent inhibitors of the M^pro. We find that the inclusion of enhanced sampling techniques such as replica-exchange algorithm in binding free energy calculations can improve the convergence of predicted non-covalent binding free energy estimates of inhibitors binding to the M^pro target. In addition, our results indicate that binding free energy calculations coupled with multiscale simulations can be a useful approach to employ in ranking covalent inhibitors to their targets. This approach may be valuable in prioritizing and refining covalent inhibitor compounds for lead discovery efforts against COVID-19 and other coronavirus infections.

Best practices for constructing, preparing, and evaluating protein-ligand binding affinity benchmarks [Article v0.1]

Free energy calculations are rapidly becoming indispensable in structure-enabled drug discovery programs. As new methods, force fields, and implementations are developed, assessing their expected accuracy on real-world systems (benchmarking) becomes critical to provide users with an assessment of the accuracy expected when these methods are applied within their domain of applicability, and developers with a way to assess the expected impact of new methodologies. These assessments require construction of a benchmark-a set of well-prepared, high quality systems with corresponding experimental measurements designed to ensure the resulting calculations provide a realistic assessment of expected performance when these methods are deployed within their domains of applicability. To date, the community has not yet adopted a common standardized benchmark, and existing benchmark reports suffer from a myriad of issues, including poor data quality, limited statistical power, and statistically deficient analyses, all of which can conspire to produce benchmarks that are poorly predictive of real-world performance. Here, we address these issues by presenting guidelines for (1) curating experimental data to develop meaningful benchmark sets, (2) preparing benchmark inputs according to best practices to facilitate widespread adoption, and (3) analysis of the resulting predictions to enable statistically meaningful comparisons among methods and force fields. We highlight challenges and open questions that remain to be solved in these areas, as well as recommendations for the collection of new datasets that might optimally serve to measure progress as methods become systematically more reliable. Finally, we provide a curated, versioned, open, standardized benchmark set adherent to these standards (PLBenchmarks) and an open source toolkit for implementing standardized best practices assessments (arsenic) for the community to use as a standardized assessment tool. While our main focus is free energy methods based on molecular simulations, these guidelines should prove useful for assessment of the rapidly growing field of machine learning methods for affinity prediction as well.

AI-Based Protein Structure Prediction in Drug Discovery: Impacts and Challenges

Proteins are the molecular machinery of the human body, and their malfunctioning is often responsible for diseases, making them crucial targets for drug discovery. The three-dimensional structure of a protein determines its biological function, its conformational state determines substrates, cofactors, and protein binding. Rational drug discovery employs engineered small molecules to selectively interact with proteins to modulate their function. To selectively target a protein and to design small molecules, knowing the protein structure with all its specific conformation is critical. Unfortunately, for a large number of proteins relevant for drug discovery, the three-dimensional structure has not yet been experimentally solved. Therefore, accurately predicting their structure based on their amino acid sequence is one of the grant challenges in biology. Recently, AlphaFold2, a machine learning application based on a deep neural network, was able to predict unknown structures of proteins with an unprecedented accuracy. Despite the impressive progress made by AlphaFold2, nature still challenges the field of structure prediction. In this Perspective, we explore how AlphaFold2 and related methods help make drug design more efficient. Furthermore, we discuss the roles of predicting domain-domain orientations, all relevant conformational states, the influence of posttranslational modifications, and conformational changes due to protein binding partners. We highlight where further improvements are needed for advanced machine learning methods to be successfully and frequently used in the pharmaceutical industry.

Alchemical free energy simulations without speed limits. A generic framework to calculate free energy differences independent of the underlying molecular dynamics program

We describe the theory of the so‐called common‐core/serial‐atom‐insertion (CC/SAI) approach to compute alchemical free energy differences and its practical implementation in a Python package called Transformato. CC/SAI is not tied to a specific biomolecular simulation program and does not rely on special purpose code for alchemical transformations. To calculate the alchemical free energy difference between several small molecules, the physical end‐states are mutated into a suitable common core. Since this only requires turning off interactions, the setup of intermediate states is straightforward to automate. Transformato currently supports CHARMM and OpenMM as back ends to carry out the necessary molecular dynamics simulations, as well as post‐processing calculations. We validate the method by computing a series of relative solvation free energy differences.

Large Scale Study of Ligand-Protein Relative Binding Free Energy Calculations: Actionable Predictions from Statistically Robust Protocols

The accurate and reliable prediction of protein-ligand binding affinities can play a central role in the drug discovery process as well as in personalized medicine. Of considerable importance during lead optimization are the alchemical free energy methods that furnish an estimation of relative binding free energies (RBFE) of similar molecules. Recent advances in these methods have increased their speed, accuracy, and precision. This is evident from the increasing number of retrospective as well as prospective studies employing them. However, such methods still have limited applicability in real-world scenarios due to a number of important yet unresolved issues. Here, we report the findings from a large data set comprising over 500 ligand transformations spanning over 300 ligands binding to a diverse set of 14 different protein targets which furnish statistically robust results on the accuracy, precision, and reproducibility of RBFE calculations. We use ensemble-based methods which are the only way to provide reliable uncertainty quantification given that the underlying molecular dynamics is chaotic. These are implemented using TIES (Thermodynamic Integration with Enhanced Sampling). Results achieve chemical accuracy in all cases. Ensemble simulations also furnish information on the statistical distributions of the free energy calculations which exhibit non-normal behavior. We find that the "enhanced sampling" method known as replica exchange with solute tempering degrades RBFE predictions. We also report definitively on numerous associated alchemical factors including the choice of ligand charge method, flexibility in ligand structure, and the size of the alchemical region including the number of atoms involved in transforming one ligand into another. Our findings provide a key set of recommendations that should be adopted for the reliable application of RBFE methods.

Estimation of binding rates and affinities from multiensemble Markov models and ligand decoupling

Accurate and efficient simulation of the thermodynamics and kinetics of protein-ligand interactions is crucial for computational drug discovery. Multiensemble Markov Model (MEMM) estimators can provide estimates of both binding rates and affinities from collections of short trajectories but have not been systematically explored for situations when a ligand is decoupled through scaling of non-bonded interactions. In this work, we compare the performance of two MEMM approaches for estimating ligand binding affinities and rates: (1) the transition-based reweighting analysis method (TRAM) and (2) a Maximum Caliber (MaxCal) based method. As a test system, we construct a small host-guest system where the ligand is a single uncharged Lennard-Jones (LJ) particle, and the receptor is an 11-particle icosahedral pocket made from the same atom type. To realistically mimic a protein-ligand binding system, the LJ ϵ parameter was tuned, and the system was placed in a periodic box with 860 TIP3P water molecules. A benchmark was performed using over 80 µs of unbiased simulation, and an 18-state Markov state model was used to estimate reference binding affinities and rates. We then tested the performance of TRAM and MaxCal when challenged with limited data. Both TRAM and MaxCal approaches perform better than conventional Markov state models, with TRAM showing better convergence and accuracy. We find that subsampling of trajectories to remove time correlation improves the accuracy of both TRAM and MaxCal and that in most cases, only a single biased ensemble to enhance sampled transitions is required to make accurate estimates.

Classical and Machine Learning Methods for Protein - Ligand Binding Free Energy Estimation

Binding free energy estimation of drug candidates to their biomolecular target is one of the best quantitative estimators in computer-aided drug discovery. Accurate binding free energy estimation is still a challengeable task even after decades of research, along with the complexity of the algorithm, time-consuming procedures, and reproducibility issues. In this review, we have discussed the advantages and disadvantages of diverse free energy methods like Thermodynamic Integration (TI), Bennett's Acceptance Ratio (BAR), Free Energy Perturbation (FEP), alchemical methods. Moreover, we discussed the possible application of the machine learning method in protein-ligand binding free energy estimation.

Hybrid QM/MM Free-Energy Evaluation of Drug-Resistant Mutational Effect on the Binding of an Inhibitor Indinavir to HIV-1 Protease

A human immunodeficiency virus-1 (HIV-1) protease is a homodimeric aspartic protease essential for the replication of HIV. The HIV-1 protease is a target protein in drug discovery for antiretroviral therapy, and various inhibitor molecules of transition state analogues have been developed. However, serious drug-resistant mutants have emerged. For understanding the molecular mechanism of the drug resistance, an accurate examination of the impacts of the mutations on ligand binding and enzymatic activity is necessary. Here, we present a molecular simulation study on the ligand binding of indinavir, a potent transition state analogue inhibitor, to the wild-type protein and a V82T/I84V drug-resistant mutant of the HIV-1 protease. We employed a hybrid ab initio quantum mechanical/molecular mechanical (QM/MM) free-energy optimization technique which combines a highly accurate QM description of the ligand molecule and its interaction with statistically ample conformational sampling of the MM protein environment by long-time molecular dynamics simulations. Through the free-energy calculations of protonation states of catalytic groups at the binding pocket and of the ligand-binding affinity changes upon the mutations, we successfully reproduced the experimentally observed significant reduction of the binding affinity upon the drug-resistant mutations and elucidated the underlying molecular mechanism. The present study opens the way for understanding the molecular mechanism of drug resistance through the direct quantitative comparison of ligand binding and enzymatic reaction with the same accuracy.

Precise Binding Free Energy Calculations for Multiple Molecules Using an Optimal Measurement Network of Pairwise Differences

Alchemical binding free energy (BFE) calculations offer an efficient and thermodynamically rigorous approach to in silico binding affinity predictions. As a result of decades of methodological improvements and recent advances in computer technology, alchemical BFE calculations are now widely used in drug discovery research. They help guide the prioritization of candidate drug molecules by predicting their binding affinities for a biomolecular target of interest (and potentially selectivity against undesirable antitargets). Statistical variance associated with such calculations, however, may undermine the reliability of their predictions, introducing uncertainty both in ranking candidate molecules and in benchmarking their predictive accuracy. Here, we present a computational method that substantially improves the statistical precision in BFE calculations for a set of ligands binding to a common receptor by dynamically allocating computational resources to different BFE calculations according to an optimality objective established in a previous work from our group and extended in this work. Our method, termed Network Binding Free Energy (NetBFE), performs adaptive BFE calculations in iterations, re-optimizing the allocations in each iteration based on the statistical variances estimated from previous iterations. Using examples of NetBFE calculations for protein binding of congeneric ligand series, we demonstrate that NetBFE approaches the optimal allocation in a small number (≤5) of iterations and that NetBFE reduces the statistical variance in the BFE estimates by approximately a factor of 2 when compared to a previously published and widely used allocation method at the same total computational cost.

SCONES: Self-Consistent Neural Network for Protein Stability Prediction Upon Mutation

Engineering proteins to have desired properties by mutating amino acids at specific sites is commonplace. Such engineered proteins must be stable to function. Experimental methods used to determine stability at throughputs required to scan the protein sequence space thoroughly are laborious. To this end, many machine learning based methods have been developed to predict thermodynamic stability changes upon mutation. These methods have been evaluated for symmetric consistency by testing with hypothetical reverse mutations. In this work, we propose transitive data augmentation, evaluating transitive consistency with our new S^transitive data set, and a new machine learning based method, the first of its kind, that incorporates both symmetric and transitive properties into the architecture. Our method, called SCONES, is an interpretable neural network that predicts small relative protein stability changes for missense mutations that do not significantly alter the structure. It estimates a residue's contributions toward protein stability (ΔG) in its local structural environment, and the difference between independently predicted contributions of the reference and mutant residues is reported as ΔΔG. We show that this self-consistent machine learning architecture is immune to many common biases in data sets, relies less on data than existing methods, is robust to overfitting, and can explain a substantial portion of the variance in experimental data.

Building intuition for binding free energy calculations: Bound state definition, restraints, and symmetry

The theory behind computation of absolute binding free energies using explicit-solvent molecular simulations is well-established, yet somewhat complex, with counter-intuitive aspects. This leads to frequent frustration, common misconceptions, and sometimes erroneous numerical treatment. To improve this, we present the main practically relevant segments of the theory with constant reference to physical intuition. We pinpoint the role of the implicit or explicit definition of the bound state (or the binding site) to make a robust link between an experimental measurement and a computational result. We clarify the role of symmetry and discuss cases where symmetry number corrections have been misinterpreted. In particular, we argue that symmetry corrections as classically presented are a source of confusion and could be advantageously replaced by restraint free energy contributions. We establish that contrary to a common intuition, partial or missing sampling of some modes of symmetric bound states does not affect the calculated decoupling free energies. Finally, we review these questions and pitfalls in the context of a few common practical situations: binding to a symmetric receptor (equivalent binding sites), binding of a symmetric ligand (equivalent poses), and formation of a symmetric complex, in the case of homodimerization.

Rapid and accurate estimation of protein-ligand relative binding affinities using site-identification by ligand competitive saturation

Predicting relative protein-ligand binding affinities is a central pillar of lead optimization efforts in structure-based drug design. The site identification by ligand competitive saturation (SILCS) methodology is based on functional group affinity patterns in the form of free energy maps that may be used to compute protein-ligand binding poses and affinities. Presented are results obtained from the SILCS methodology for a set of eight target proteins as reported originally in Wang et al. (J. Am. Chem. Soc., 2015, 137, 2695-2703) using free energy perturbation (FEP) methods in conjunction with enhanced sampling and cycle closure corrections. These eight targets have been subsequently studied by many other authors to compare the efficacy of their method while comparing with the outcomes of Wang et al. In this work, we present results for a total of 407 ligands on the eight targets and include specific analysis on the subset of 199 ligands considered previously. Using the SILCS methodology we can achieve an average accuracy of up to 77% and 74% when considering the eight targets with their 199 and 407 ligands, respectively, for rank-ordering ligand affinities as calculated by the percent correct metric. This accuracy increases to 82% and 80%, respectively, when the SILCS atomic free energy contributions are optimized using a Bayesian Markov-chain Monte Carlo approach. We also report other metrics including Pearson's correlation coefficient, Pearlman's predictive index, mean unsigned error, and root mean square error for both sets of ligands. The results obtained for the 199 ligands are compared with the outcomes of Wang et al. and other published works. Overall, the SILCS methodology yields similar or better-quality predictions without a priori need for known ligand orientations in terms of the different metrics when compared to current FEP approaches with significant computational savings while additionally offering quantitative estimates of individual atomic contributions to binding free energies. These results further validate the SILCS methodology as an accurate, computationally efficient tool to support lead optimization and drug discovery.

Robust and Efficient Implicit Solvation Model for Fast Semiempirical Methods

We present a robust and efficient method to implicitly account for solvation effects in modern semiempirical quantum mechanics and force fields. A computationally efficient yet accurate solvation model based on the analytical linearized Poisson-Boltzmann (ALPB) model is parameterized for the extended tight binding (xTB) and density functional tight binding (DFTB) methods as well as for the recently proposed GFN-FF general force field. The proposed methods perform well over a broad range of systems and applications, from conformational energies over transition-metal complexes to large supramolecular association reactions of charged species. For hydration free energies of small molecules, GFN1-xTB(ALPB) is reaching the accuracy of sophisticated explicitly solvated approaches, with a mean absolute deviation of only 1.4 kcal/mol compared to the experiment. Logarithmic octanol-water partition coefficients (log K_ow) are computed with a mean absolute deviation of about 0.65 using GFN2-xTB(ALPB) compared to experimental values indicating a consistent description of differential solvent effects. Overall, more than twenty solvents for each of the six semiempirical methods are parameterized and tested. They are readily available in the xtb and dftb+ programs for diverse computational applications.

Dynamic Cooperativity of Ligand-Residue Interactions Evaluated with the Fragment Molecular Orbital Method

By the splendid advance in computation power realized with the Fugaku supercomputer, it has become possible to perform ab initio fragment molecular orbital (FMO) calculations for thousands of dynamic structures of protein-ligand complexes in a parallel way. We thus carried out electron-correlated FMO calculations for a complex of the 3C-like (3CL) main protease (M^pro) of the new coronavirus (SARS-CoV-2) and its inhibitor N3 incorporating the structural fluctuations sampled by classical molecular dynamics (MD) simulation in hydrated conditions. Along with a statistical evaluation of the interfragment interaction energies (IFIEs) between the N3 ligand and the surrounding amino-acid residues for 1000 dynamic structure samples, in this study we applied a novel approach based on principal component analysis (PCA) and singular value decomposition (SVD) to the analysis of IFIE data in order to extract the dynamically cooperative interactions between the ligand and the residues. We found that the relative importance of each residue is modified via the structural fluctuations and that the ligand is bound in the pharmacophore in a dynamic manner through collective interactions formed by multiple residues, thus providing new insight into structure-based drug discovery.

Scaffold Hopping Transformations Using Auxiliary Restraints for Calculating Accurate Relative Binding Free Energies

In silico screening of drug-target interactions is a key part of the drug discovery process. Changes in the drug scaffold via contraction or expansion of rings, the breaking of rings, and the introduction of cyclic structures from acyclic structures are commonly applied by medicinal chemists to improve binding affinity and enhance favorable properties of candidate compounds. These processes, commonly referred to as scaffold hopping, are challenging to model computationally. Although relative binding free energy (RBFE) calculations have shown success in predicting binding affinity changes caused by perturbing R-groups attached to a common scaffold, applications of RBFE calculations to modeling scaffold hopping are relatively limited. Scaffold hopping inevitably involves breaking and forming bond interactions of quadratic functional forms, which is highly challenging. A novel method for handling ring opening/closure/contraction/expansion and linker contraction/expansion is presented here. To the best of our knowledge, RBFE calculations on linker contraction/expansion have not been previously reported. The method uses auxiliary restraints to hold the atoms at the ends of a bond in place during the breaking and forming of the bonds. The broad applicability of the method was demonstrated by examining perturbations involving small-molecule macrocycles and mutations of proline in proteins. High accuracy was obtained using the method for most of the perturbations studied. The rigor of the method was isolated from the force field by validating the method using relative and absolute hydration free energy calculations compared to standard simulation results. Unlike other methods that rely on λ-dependent functional forms for bond interactions, the method presented here can be employed using modern molecular dynamics software without modification of codes or force field functions.

A Cloud Computing Platform for Scalable Relative and Absolute Binding Free Energy Predictions: New Opportunities and Challenges for Drug Discovery

Free energy perturbation (FEP) has become widely used in drug discovery programs for binding affinity prediction between candidate compounds and their biological targets. However, limitations of FEP applications also exist, including, but not limited to, high cost, long waiting time, limited scalability, and breadth of application scenarios. To overcome these problems, we have developed XFEP, a scalable cloud computing platform for both relative and absolute free energy predictions using optimized simulation protocols. XFEP enables large-scale FEP calculations in a more efficient, scalable, and affordable way, for example, the evaluation of 5000 compounds can be performed in 1 week using 50-100 GPUs with a computing cost roughly equivalent to the cost for the synthesis of only one new compound. By combining these capabilities with artificial intelligence techniques for goal-directed molecule generation and evaluation, new opportunities can be explored for FEP applications in the drug discovery stages of hit identification, hit-to-lead, and lead optimization based not only on structure exploitation within the given chemical series but also including evaluation and comparison of completely unrelated molecules during structure exploration in a larger chemical space. XFEP provides the basis for scalable FEP applications to become more widely used in drug discovery projects and to speed up the drug discovery process from hit identification to preclinical candidate compound nomination.

Artificial intelligence in drug design: algorithms, applications, challenges and ethics

The discovery paradigm of drugs is rapidly growing due to advances in machine learning (ML) and artificial intelligence (AI). This review covers myriad faces of AI and ML in drug design. There is a plethora of AI algorithms, the most common of which are summarized in this review. In addition, AI is fraught with challenges that are highlighted along with plausible solutions to them. Examples are provided to illustrate the use of AI and ML in drug discovery and in predicting drug properties such as binding affinities and interactions, solubility, toxicology, blood–brain barrier permeability and chemical properties. The review also includes examples depicting the implementation of AI and ML in tackling intractable diseases such as COVID-19, cancer and Alzheimer’s disease. Ethical considerations and future perspectives of AI are also covered in this review.

Challenges Encountered Applying Equilibrium and Nonequilibrium Binding Free Energy Calculations

Binding free energy calculations have become increasingly valuable to drive decision making in drug discovery projects. However, among other issues, inadequate sampling can reduce accuracy, limiting the value of the technique. In this paper, we apply absolute binding free energy calculations to ligands binding to T4 lysozyme L99A and HSP90 using equilibrium and nonequilibrium approaches. We highlight sampling problems encountered in these systems, such as slow side chain rearrangements and slow changes of water placement upon ligand binding. These same types of challenges are also likely to show up in other protein-ligand systems, and we propose some strategies to diagnose and test for such problems in alchemical free energy calculations. We also explore similarities and differences in how the equilibrium and the nonequilibrium approaches handle these problems. Our results show the large amount of work still to be done to make free energy calculations robust and reliable and provide insight for future research in this area.

Conformational Shifts of Stacked Heteroaromatics: Vacuum vs. Water Studied by Machine Learning

Stacking interactions play a crucial role in drug design, as we can find aromatic cores or scaffolds in almost any available small molecule drug. To predict optimal binding geometries and enhance stacking interactions, usually high-level quantum mechanical calculations are performed. These calculations have two major drawbacks: they are very time consuming, and solvation can only be considered using implicit solvation. Therefore, most calculations are performed in vacuum. However, recent studies have revealed a direct correlation between the desolvation penalty, vacuum stacking interactions and binding affinity, making predictions even more difficult. To overcome the drawbacks of quantum mechanical calculations, in this study we use neural networks to perform fast geometry optimizations and molecular dynamics simulations of heteroaromatics stacked with toluene in vacuum and in explicit solvation. We show that the resulting energies in vacuum are in good agreement with high-level quantum mechanical calculations. Furthermore, we show that using explicit solvation substantially influences the favored orientations of heteroaromatic rings thereby emphasizing the necessity to include solvation properties starting from the earliest phases of drug design.

Implementing and Assessing an Alchemical Method for Calculating Protein-Protein Binding Free Energy

Protein-protein binding is fundamental to most biological processes. It is important to be able to use computation to accurately estimate the change in protein-protein binding free energy due to mutations in order to answer biological questions that would be experimentally challenging, laborious, or time-consuming. Although nonrigorous free-energy methods are faster, rigorous alchemical molecular dynamics-based methods are considerably more accurate and are becoming more feasible with the advancement of computer hardware and molecular simulation software. Even with sufficient computational resources, there are still major challenges to using alchemical free-energy methods for protein-protein complexes, such as generating hybrid structures and topologies, maintaining a neutral net charge of the system when there is a charge-changing mutation, and setting up the simulation. In the current study, we have used the pmx package to generate hybrid structures and topologies, and a double-system/single-box approach to maintain the net charge of the system. To test the approach, we predicted relative binding affinities for two protein-protein complexes using a nonequilibrium alchemical method based on the Crooks fluctuation theorem and compared the results with experimental values. The method correctly identified stabilizing from destabilizing mutations for a small protein-protein complex, and a larger, more challenging antibody complex. Strong correlations were obtained between predicted and experimental relative binding affinities for both protein-protein systems.

Qualitative Prediction of Ligand Dissociation Kinetics from Focal Adhesion Kinase Using Steered Molecular Dynamics

Most early-stage drug discovery projects focus on equilibrium binding affinity to the target alongside selectivity and other pharmaceutical properties. Since many approved drugs have nonequilibrium binding characteristics, there has been increasing interest in optimizing binding kinetics early in the drug discovery process. As focal adhesion kinase (FAK) is an important drug target, we examine whether steered molecular dynamics (SMD) can be useful for identifying drug candidates with the desired drug-binding kinetics. In simulating the dissociation of 14 ligands from FAK, we find an empirical power–law relationship between the simulated time needed for ligand unbinding and the experimental rate constant for dissociation, with a strong correlation depending on the SMD force used. To improve predictions, we further develop regression models connecting experimental dissociation rate with various structural and energetic quantities derived from the simulations. These models can be used to predict dissociation rates from FAK for related compounds.