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.

On computing the solubility of molecular systems subject to constraints using the extended Einstein crystal method.

A method to compute solubilities for molecular systems using atomistic simulations, based on an extension of the Einstein crystal method, has recently been presented [Li et al., J. Chem. Phys. 146, 214110 (2017)]. This methodology is particularly appealing to compute solubilities in cases of practical importance including, but not limited to, solutions where the solute is sparingly soluble and molecules of importance for the pharmaceutical industry, which are often characterized by strong polar interactions and slow relaxation time scales. The mathematical derivation of this methodology hinges on a factorization of the partition function which is not necessarily applicable in the case of a system subject to holonomic molecular constraints. We show here that, although the mathematical procedure to derive it is slightly different, essentially the same mathematical relation for calculating the solubility can be safely applied for computing the solubility of systems subject to constraints, which are the majority of the systems used for practical molecular simulations.

Solubility of paracetamol in ethanol by molecular dynamics using the extended Einstein crystal method and experiments.

Li and co-workers [Li et al., J. Chem. Phys. 146, 214110 (2017)] have recently proposed a methodology to compute the solubility of molecular compounds from first principles, using molecular dynamics simulations. We revise and further explore their methodology that was originally applied to naphthalene in water at low concentration. In particular, we compute the solubility of paracetamol in an ethanol solution at ambient conditions. For the simulations, we used a force field that we previously reparameterized to reproduce certain thermodynamic properties of paracetamol but not explicitly its solubility in ethanol. In addition, we have determined the experimental solubility by performing turbidity measurements using a Crystal16 over a range of temperatures. Our work serves a dual purpose: (i) methodologically, we clarify how to compute, with a relatively straightforward procedure, the solubility of molecular compounds and (ii) applying this procedure, we show that the solubility predicted by our force field (0.085 ± 0.014 in mole ratio) is in good agreement with the experimental value obtained from our experiments and those reported in the literature (average 0.0585 ± 0.004), considering typical deviations for predictions from first principle methods. The good agreement between the experimental and the calculated solubility also suggests that the method used to reparameterize the force field can be used as a general strategy to optimize force fields for simulations in solution.

Machine Learning Nucleation Collective Variables with Graph Neural Networks.

The efficient calculation of nucleation collective variables (CVs) is one of the main limitations to the application of enhanced sampling methods to the investigation of nucleation processes in realistic environments. Here we discuss the development of a graph-based model for the approximation of nucleation CVs that enables orders-of-magnitude gains in computational efficiency in the on-the-fly evaluation of nucleation CVs. By performing simulations on a nucleating colloidal system mimicking a multistep nucleation process from solution, we assess the model's efficiency in both postprocessing and on-the-fly biasing of nucleation trajectories with pulling, umbrella sampling, and metadynamics simulations. Moreover, we probe and discuss the transferability of graph-based models of nucleation CVs across systems using the model of a CV based on sixth-order Steinhardt parameters trained on a colloidal system to drive the nucleation of crystalline copper from its melt. Our approach is general and potentially transferable to more complex systems as well as to different CVs.

Development and Comprehensive Benchmark of a High-Quality AMBER-Consistent Small Molecule Force Field with Broad Chemical Space Coverage for Molecular Modeling and Free Energy Calculation.

Biomolecular simulations have become an essential tool in contemporary drug discovery, and molecular mechanics force fields (FFs) constitute its cornerstone. Developing a high quality and broad coverage general FF is a significant undertaking that requires substantial expert knowledge and computing resources, which is beyond the scope of general practitioners. Existing FFs originate from only a limited number of groups and organizations, and they either suffer from limited numbers of training sets, lower than desired quality because of oversimplified representations, or are costly for the molecular modeling community to access. To address these issues, in this work, we developed an AMBER-consistent small molecule FF with extensive chemical space coverage, and we provide Open Access parameters for the entire modeling community. To validate our FF, we carried out benchmarks of quantum mechanics (QM)/molecular mechanics conformer comparison and free energy perturbation calculations on several benchmark data sets. Our FF achieves a higher level of performance at reproducing QM energies and geometries than two popular open-source FFs, OpenFF2 and GAFF2. In relative binding free energy calculations for 31 protein-ligand data sets, comprising 1079 pairs of ligands, the new FF achieves an overall root-mean-square error of 1.19 kcal/mol for ΔΔG and 0.92 kcal/mol for ΔG on a subset of 463 ligands without bespoke fitting to the data sets. The results are on par with those of the leading commercial series of OPLS FFs.

Absolute transition rates for rare events from dynamical decoupling of reaction variables.

We introduce a new approach to evaluate transition rates for rare events in complex many-particle systems. Building on a path-integral representation of transition probabilities for Markov processes, the rate is first expressed in terms of a free energy in the transition-path ensemble. We then define an auxiliary process where a suitably defined reaction variable is dynamically decoupled from all the others, whose dynamics is left unchanged. For this system the transition rates coincide with those of a unidimensional process whose only coordinate is the reaction variable. The free-energy difference between the auxiliary and the physical transition-path ensembles is finally evaluated using standard techniques. The efficiency of this method is deemed to be optimal because the physical and auxiliary dynamics differ by one degree of freedom only at any system size. Our method is demonstrated numerically on a simple model of Lennard-Jones particles ruled by the overdamped Langevin equation.

Computational Modeling of the Disulfide Cross-Linking Reaction.

Disulfide cross-linking is one of the fundamental covalent bonds that exist prevalently in many biological molecules that is involved in versatile functional activities such as antibody stability, viral assembly, and protein folding. Additionally, it is a crucial factor in various industrial applications. Therefore, a fundamental understanding of its reaction mechanism would help gain insight into its different functional activities. Computational simulation of the disulfide cross-linking reaction with hydrogen peroxide (H2O2) was performed at the integrated quantum mechanical/molecular mechanical (QM/MM) level of theory in a water box under periodic boundary conditions. A benchmarking study for the barrier height of the disulfide formation step was performed on a model system between methanethiol and methane sulfenic acid to determine, for the QM system, the best-fit density functional theory (DFT) functional/basis set combination that produces comparable results to a higher-level theory of the coupled-cluster method. Computational results show that the disulfide cross-linking reaction with H2O2 reagent can proceed through a one-step or a two-step pathway for the high pKa cysteines or two different pathways for the low pKa cysteines to ultimately produce the sulfenic acid/sulfenate intermediate complex. Subsequently, those intermediates react with another neutral/anionic cysteine residue to form the cysteine product. In addition, the solvent-assisted proton-exchange/proton-transfer effects were examined on the energetic barriers for the different transition states, and the molecular contributions of the chemically involved water molecules were studied in detail.

Nucleation of Molecular Crystals Driven by Relative Information Entropy.

Simulating nucleation of molecular crystals is extremely challenging for all but the simplest cases. The challenge lies in formulating effective order parameters that are capable of driving the transition process. In recent years, order parameters based on molecular pair-functions have been successfully used in combination with enhanced sampling techniques to simulate nucleation of simple molecular crystals. However, despite the success of these approaches, we demonstrate that they can fail when applied to more complex cases. In fact, we show that order parameters based on molecular pair-functions, while successful at nucleating benzene, fail for paracetamol. Hence, we introduce a novel approach to formulate order parameters. In our approach, we construct reduced dimensional distributions of relevant quantities on the fly and then quantify the difference between these distributions and selected reference distributions. By computing the distribution of different quantities and by choosing different reference distributions, it is possible to systematically construct an effective set of order parameters. We then show that our new order parameters are capable of driving the nucleation of ordered states and, in particular, the form I crystal of paracetamol.

Extended Hamiltonian approach to continuous tempering.

We introduce an enhanced sampling simulation technique based on continuous tempering, i.e., on continuously varying the temperature of the system under investigation. Our approach is mathematically straightforward, being based on an extended Hamiltonian formulation in which an auxiliary degree of freedom, determining the effective temperature, is coupled to the physical system. The physical system and its temperature evolve continuously in time according to the equations of motion derived from the extended Hamiltonian. Due to the Hamiltonian structure, it is easy to show that a particular subset of the configurations of the extended system is distributed according to the canonical ensemble for the physical system at the correct physical temperature.

Investigating Molecular Kinetics by Variationally Optimized Diffusion Maps.

Identification of the collective coordinates that describe rare events in complex molecular transitions such as protein folding has been a key challenge in the theoretical molecular sciences. In the Diffusion Map approach, one assumes that the molecular configurations sampled have been generated by a diffusion process, and one uses the eigenfunctions of the corresponding diffusion operator as reaction coordinates. While diffusion coordinates (DCs) appear to provide a good approximation to the true dynamical reaction coordinates, they are not parametrized using dynamical information. Thus, their approximation quality could not, as yet, be validated, nor could the diffusion map eigenvalues be used to compute relaxation rate constants of the system. Here we combine the Diffusion Map approach with the recently proposed Variational Approach for Conformation Dynamics (VAC). Diffusion Map coordinates are used as a basis set, and their optimal linear combination is sought using the VAC, which employs time-correlation information on the molecular dynamics (MD) trajectories. We have applied this approach to ultra-long MD simulations of the Fip35 WW domain and found that the first DCs are indeed a good approximation to the true reaction coordinates of the system, but they could be further improved using the VAC. Using the Diffusion Map basis, excellent approximations to the relaxation rates of the system are obtained. Finally, we evaluate the quality of different metric spaces and find that pairwise minimal root-mean-square deviation performs poorly, while operating in the recently introduced kinetic maps based on the time-lagged independent component analysis gives the best performance.