Deep learning has disrupted nearly every field of research, including those of direct importance to drug discovery, such as medicinal chemistry and pharmacology. This revolution has largely been ...attributed to the unprecedented advances in highly parallelizable graphics processing units (GPUs) and the development of GPU-enabled algorithms. In this Review, we present a comprehensive overview of historical trends and recent advances in GPU algorithms and discuss their immediate impact on the discovery of new drugs and drug targets. We also cover the state-of-the-art of deep learning architectures that have found practical applications in both early drug discovery and consequent hit-to-lead optimization stages, including the acceleration of molecular docking, the evaluation of off-target effects and the prediction of pharmacological properties. We conclude by discussing the impacts of GPU acceleration and deep learning models on the global democratization of the field of drug discovery that may lead to efficient exploration of the ever-expanding chemical universe to accelerate the discovery of novel medicines.GPUs, which are highly parallel computer processing units, were originally designed for graphics applications, but they have played an important role in accelerating the development of deep learning methods. In this Review, Pandey and colleagues summarize how GPUs have advanced machine learning in the field of drug discovery.
Monte Carlo simulations were performed modeling hydrogen sorption in a recently synthesized metal−organic framework material (MOF) that exhibits large molecular hydrogen uptake capacity. The MOF is ...remarkable because at 78 K and 1.0 atm it sorbs hydrogen at a density near that of liquid hydrogen (at 20 K and 1.0 atm) when considering H2 density in the pores. Unlike most other MOFs that have been investigated for hydrogen storage, it has a highly ionic framework and many relatively small channels. The simulations demonstrate that it is both of these physical characteristics that lead to relatively strong hydrogen interactions in the MOF and ultimately large hydrogen uptake. Microscopically, hydrogen interacts with the MOF via three principle attractive potential energy contributions: Van der Waals, charge-quadrupole, and induction. Previous simulations of hydrogen storage in MOFs and other materials have not focused on the role of polarization effects, but they are demonstrated here to be the dominant contribution to hydrogen physisorption. Indeed, polarization interactions in the MOF lead to two distinct populations of dipolar hydrogen that are identified from the simulations that should be experimentally discernible using, for example, Raman spectroscopy. Since polarization interactions are significantly enhanced by the presence of a charged framework with narrow pores, MOFs are excellent hydrogen storage candidates.
We develop a generalizable AI-driven workflow that leverages heterogeneous HPC resources to explore the time-dependent dynamics of molecular systems. We use this workflow to investigate the ...mechanisms of infectivity of the SARS-CoV-2 spike protein, the main viral infection machinery. Our workflow enables more efficient investigation of spike dynamics in a variety of complex environments, including within a complete SARS-CoV-2 viral envelope simulation, which contains 305 million atoms and shows strong scaling on ORNL Summit using NAMD. We present several novel scientific discoveries, including the elucidation of the spike’s full glycan shield, the role of spike glycans in modulating the infectivity of the virus, and the characterization of the flexible interactions between the spike and the human ACE2 receptor. We also demonstrate how AI can accelerate conformational sampling across different systems and pave the way for the future application of such methods to additional studies in SARS-CoV-2 and other molecular systems.
It is now well established by numerous experimental and computational studies that the adsorption propensities of inorganic anions conform to the Hofmeister series. The adsorption propensities of ...inorganic cations, such as the alkali metal cations, have received relatively little attention. Here we use a combination of liquid-jet X-ray photoelectron experiments and molecular dynamics simulations to investigate the behavior of K⁺ and Li⁺ ions near the interfaces of their aqueous solutions with halide ions. Both the experiments and the simulations show that Li⁺ adsorbs to the aqueous solution–vapor interface, while K⁺ does not. Thus, we provide experimental validation of the “surfactant-like” behavior of Li⁺ predicted by previous simulation studies. Furthermore, we use our simulations to trace the difference in the adsorption of K⁺ and Li⁺ ions to a difference in the resilience of their hydration shells.
We report molecular dynamics simulations of acetonitrile–water binary solutions at concentrations of 0.032–0.59 mole fraction. We find that at low bulk concentration acetonitrile has an enhanced ...population near the liquid/vapor interface. The surface-bound acetonitrile molecules exhibit anisotropic orientations and lie nearly flat along the solution surface with their terminal methyl groups directed toward the vapor. Upon increasing the bulk concentration, the formation of acetonitrile domains is promoted by interactions between hydrophobic methyl moieties. Dipole–dipole interactions facilitate a pseudonematic, antiparallel pairing of near-neighbor molecules both in the bulk solution and near the liquid/vapor interface. Near the interface the preferred orientation of acetonitrile flattens further to accommodate antiparallel pairing of neighboring molecules such that the methyl group remains above the solution. This study paints a surprisingly complex picture of a binary organic–water solution that manifests behavior similar to liquid crystals through preferred orientations and pseudonematic antiparallel pairing.
Nitric acid is a prevalent component of atmospheric aerosols, and the extent of nitric acid dissociation at aqueous interfaces is relevant to its role in heterogeneous atmospheric chemistry. Several ...experimental and theoretical studies have suggested that the extent of dissociation of nitric acid near aqueous interfaces is less than that in bulk solution. Here dissociation of HNO3 at the surface of aqueous solution is quantified using X-ray photoelectron spectroscopy of the nitrogen local electronic structure. The relative amounts of undissociated HNO3(aq) and dissociated NO3 –(aq) are identified by the distinguishable N1s core-level photoelectron spectra of the two species, and we determine the degree of dissociation, αint, in the interface (approximately the first three layers of solution) as a function of HNO3 concentration. Our measurements show that dissociation is decreased by ∼20% near the solution interface compared with bulk solution and furthermore that dissociation occurs in the topmost solution layer. The experimental results are supported by first-principles MD simulations, which show that hydrogen bonds between HNO3 and water molecules at the solution surface stabilize the molecular form even at low concentration by analogy to the stabilization of molecular HNO3 that occurs in bulk solution at high concentration.
An anisotropic many-body H2 potential energy function has been developed for use in heterogeneous systems. The intermolecular potential has been derived from first principles and expressed in a form ...that is readily transferred to exogenous systems, e.g. in modeling H2 sorption in solid-state materials. Explicit many-body polarization effects, known to be important in simulating hydrogen at high density, are incorporated. The analytic form of the potential energy function is suitable for methods of statistical physics, such as Monte Carlo or Molecular Dynamics simulation. The model has been validated on dense supercritical hydrogen and demonstrated to reproduce the experimental data with high accuracy.
We present a method for fitting atomic charges to the electrostatic potential (ESP) of periodic and nonperiodic systems. This method is similar to the method of Campañá et al. J. Chem. Theory ...Comput. 2009, 5, 2866 . We compare the Wolf and Ewald long-range electrostatic summation methods in calculating the ESP for periodic systems. We find that the Wolf summation is computationally more efficient than the Ewald summation by about a factor of 5 with comparable accuracy. Our analysis shows that the choice of grid mesh size influences the fitted atomic charges, especially for systems with buried (highly coordinated) atoms. We find that a maximum grid spacing of 0.2−0.3 Å is required to obtain reliable atomic charges. The effect of the exclusion radius for point selection is assessed; we find that the common choice of using the van der Waals (vdW) radius as the exclusion radius for each atom may result in large deviations between the ESP generated from the ab initio calculations and that computed from the fitted charges, especially for points closest to the exclusion radii. We find that a larger value of exclusion radius than commonly used, 1.3 times the vdW radius, provides more reliable results. We find that a penalty function approach for fitting charges for buried atoms, with the target charge taken from Bader charge analysis, gives physically reasonable results.
We demonstrate that the driving forces for ion adsorption to the air–water interface for point charge models result from both cavitation and a term that is of the form of a negative electrochemical ...surface potential. We carefully characterize the role of the free energy due to the electrochemical surface potential computed from simple empirical models and its role in ionic adsorption within the context of dielectric continuum theory. Our research suggests that the electrochemical surface potential due to point charge models provides anions with a significant driving force for adsoprtion to the air–water interface. This is contrary to the results of ab initio simulations that indicate that the average electrostatic surface potential should favor the desorption of anions at the air–water interface. The results have profound implications for the studies of ionic distributions in the vicinity of hydrophobic surfaces and proteins.