Chemical Space Exploration with Active Learning and Alchemical Free Energies.

Drug discovery can be thought of as a search for a needle in a haystack: searching through a large chemical space for the most active compounds. Computational techniques can narrow the search space for experimental follow up, but even they become unaffordable when evaluating large numbers of molecules. Therefore, machine learning (ML) strategies are being developed as computationally cheaper complementary techniques for navigating and triaging large chemical libraries. Here, we explore how an active learning protocol can be combined with first-principles based alchemical free energy calculations to identify high affinity phosphodiesterase 2 (PDE2) inhibitors. We first calibrate the procedure using a set of experimentally characterized PDE2 binders. The optimized protocol is then used prospectively on a large chemical library to navigate toward potent inhibitors. In the active learning cycle, at every iteration a small fraction of compounds is probed by alchemical calculations and the obtained affinities are used to train ML models. With successive rounds, high affinity binders are identified by explicitly evaluating only a small subset of compounds in a large chemical library, thus providing an efficient protocol that robustly identifies a large fraction of true positives.

Non-equilibrium approach for binding free energies in cyclodextrins in SAMPL7: force fields and software.

In the current work we report on our participation in the SAMPL7 challenge calculating absolute free energies of the host-guest systems, where 2 guest molecules were probed against 9 hosts-cyclodextrin and its derivatives. Our submission was based on the non-equilibrium free energy calculation protocol utilizing an averaged consensus result from two force fields (GAFF and CGenFF). The submitted prediction achieved accuracy of [Formula: see text] in terms of the unsigned error averaged over the whole dataset. Subsequently, we further report on the underlying reasons for discrepancies between our calculations and another submission to the SAMPL7 challenge which employed a similar methodology, but disparate ligand and water force fields. As a result we have uncovered a number of issues in the dihedral parameter definition of the GAFF 2 force field. In addition, we identified particular cases in the molecular topologies where different software packages had a different interpretation of the same force field. This latter observation might be of particular relevance for systematic comparisons of molecular simulation software packages. The aforementioned factors have an influence on the final free energy estimates and need to be considered when performing alchemical calculations.

Accurate absolute free energies for ligand-protein binding based on non-equilibrium approaches.

The accurate calculation of the binding free energy for arbitrary ligand-protein pairs is a considerable challenge in computer-aided drug discovery. Recently, it has been demonstrated that current state-of-the-art molecular dynamics (MD) based methods are capable of making highly accurate predictions. Conventional MD-based approaches rely on the first principles of statistical mechanics and assume equilibrium sampling of the phase space. In the current work we demonstrate that accurate absolute binding free energies (ABFE) can also be obtained via theoretically rigorous non-equilibrium approaches. Our investigation of ligands binding to bromodomains and T4 lysozyme reveals that both equilibrium and non-equilibrium approaches converge to the same results. The non-equilibrium approach achieves the same level of accuracy and convergence as an equilibrium free energy perturbation (FEP) method enhanced by Hamiltonian replica exchange. We also compare uni- and bi-directional non-equilibrium approaches and demonstrate that considering the work distributions from both forward and reverse directions provides substantial accuracy gains. In summary, non-equilibrium ABFE calculations are shown to yield reliable and well-converged estimates of protein-ligand binding affinity.

Improved general-purpose five-point model for water: TIP5P/2018.

A new five point potential for liquid water, TIP5P/2018, is presented along with the techniques used to derive its charges from ab initio per-molecule electrostatic potentials in the liquid phase using the split charge equilibration of Nistor et al. [J. Chem. Phys. 125, 094108 (2006)]. By taking the density and diffusion dependence on temperature as target properties, significant improvements to the behavior of isothermal compressibility were achieved along with improvements to other thermodynamic and rotational properties. While exhibiting a dipole moment close to ab initio values, TIP5P/2018 suffers from a too small quadrupole moment due to the charge assignment procedure and results in an overestimation of the dielectric constant.

Electronic Excitations in Complex Molecular Environments: Many-Body Green's Functions Theory in VOTCA-XTP.

Many-body Green's functions theory within the GW approximation and the Bethe-Salpeter Equation (BSE) is implemented in the open-source VOTCA-XTP software, aiming at the calculation of electronically excited states in complex molecular environments. Based on Gaussian-type atomic orbitals and making use of resolution of identity techniques, the code is designed specifically for nonperiodic systems. Application to a small molecule reference set successfully validates the methodology and its implementation for a variety of excitation types covering an energy range from 2 to 8 eV in single molecules. Further, embedding each GW-BSE calculation into an atomistically resolved surrounding, typically obtained from Molecular Dynamics, accounts for effects originating from local fields and polarization. Using aqueous DNA as a prototypical system, different levels of electrostatic coupling between the regions in this GW-BSE/MM setup are demonstrated. Particular attention is paid to charge-transfer (CT) excitations in adenine base pairs. It is found that their energy is extremely sensitive to the specific environment and to polarization effects. The calculated redshift of the CT excitation energy compared to a nucelobase dimer treated in vacuum is of the order of 1 eV, which matches expectations from experimental data. Predicted lowest CT energies are below that of a single nucleobase excitation, indicating the possibility of an initial (fast) decay of such an UV excited state into a binucleobase CT exciton. The results show that VOTCA-XTP's GW-BSE/MM is a powerful tool to study a wide range of types of electronic excitations in complex molecular environments.

Non-conformal coarse-grained potentials for water.

Water is a notoriously difficult substance to model both accurately and efficiently. Here, we focus on descriptions with a single coarse-grained particle per molecule using the so-called approximate non-conformal and generalized Stockmayer potentials as the starting points. They are fitted using the radial distribution function and the liquid-gas density profile of the atomistic extended simple point charge (SPC/E) model by downhill simplex optimization. We compare the results with monatomic water (mW), ELBA, and direct iterative Boltzmann inversion of SPC/E. The results show that symmetrical potentials result in non-transferable models, that is, they need to be reparametrized for new state points. This indicates that transferability may require more complex models. Furthermore, the results also show that the addition of a point dipole is not sufficient to make the potentials accurate and transferable to different temperatures (300 K-500 K) and pressures without an appropriate choice of properties as targets during model optimization.

Redefining the minimal substrate tolerance of mandelate racemase. Racemization of trifluorolactate.

Mandelate racemase (EC 5.1.2.2) from Pseudomonas putida catalyzes the interconversion of the enantiomers of mandelic acid and a variety of aryl- and heteroaryl-substituted mandelate derivatives, suggesting that ß,γ-unsaturation is a requisite feature of substrates for the enzyme. We show that ß,γ-unsaturation is not an absolute requirement for catalysis and that mandelate racemase can bind and catalyze the racemization of (S)-trifluorolactate (k(cat) = 2.5 ± 0.3 s(-1), K(m) = 1.74 ± 0.08 mM) and (R)-trifluorolactate (k(cat) = 2.0 ± 0.2 s(-1), K(m) = 1.2 ± 0.2 mM). The enzyme was shown to catalyze hydrogen-deuterium exchange at the α-postion of trifluorolactate using (1)H NMR spectrocsopy. ß-Elimination of fluoride was not detected using (19)F NMR spectroscopy. Although mandelate racemase bound trifluorolactate with an affinity similar to that exhibited for mandelate, the turnover numbers (k(cat)) were markedly reduced by ∼318-fold, resulting in catalytic efficiencies (k(cat)/K(m)) that were ~400-fold lower than those observed for mandelate. These observations suggested that chemical steps on the enzyme were likely rate-determining, which was confirmed by demonstrating that the rates of mandelate racemase-catalyzed racemization of (S)-trifluorolactate were not dependent upon the solvent microviscosity. Circular dichroism spectroscopy was used to measure the rates of nonenzymatic racemization of (S)-trifluorolactate at elevated temperatures. The values of ΔH(‡) and ΔS(‡) for the nonenzymatic racemization reaction were determined to be 28.0 (±0.7) kcal/mol and -15.7 (±1.7) cal K(-1) mol(-1), respectively, corresponding to a free energy of activation equal to 33 (±4) kcal/mol at 25 °C. Hence, mandelate racemase stabilizes the altered trifluorolactate in the transition state (ΔG(tx)) by at least 20 kcal/mol.