No abstract

In this manuscript, we develop multiple machine learning (ML) models to accelerate a scheme for parameterizing site-based models of exciton dynamics from all-atom configurations of condensed phase sexithiophene systems. This scheme encodes the details of a system’s specific molecular morphology in the correlated distributions of model parameters through the analysis of many single-molecule excited-state electronic-structure calculations. These calculations yield excitation energies for each molecule in the system and the network of pair-wise intermolecular electronic couplings. Here, we demonstrate that the excitation energies can be accurately predicted using a kernel ridge regression (KRR) model with Coulomb matrix featurization. We present two ML models for predicting intermolecular couplings. The first one utilizes a deep neural network and bi-molecular featurization to predict the coupling directly, which we find to perform poorly. The second one utilizes a KRR model to predict unimolecular transition densities, which can subsequently be analyzed to compute the coupling. We find that the latter approach performs excellently, indicating that an effective, generalizable strategy for predicting simple bimolecular properties is through the indirect application of ML to predict higher-order unimolecular properties. Such an approach necessitates a much smaller feature space and can incorporate the insight of well-established molecular physics.

The standard approach to calculating the dielectric constant from molecular dynamics (MD) simulations employs a variant of the Kirkwood–Fröhlich methodology. Many popular nonpolarizable models of water, such as TIPnP, give a reasonable agreement with the experimental value of 78. However, it has been argued in the literature that the dipole moments of these models are effective, being smaller than the real dipole of a liquid water molecule by about a factor of , or roughly . If the total or corrected dipole moment is used in calculations, the dielectric constant comes out nearly twice as large, i.e., in the range of 160, which is twice as high as the experimental value. Here we discuss possible reasons for such a discrepancy. One approach takes into account dynamic corrections due to the dependence of the dielectric response of the medium producing the reaction field on the time scale of dipole fluctuations computed in the Kirkwood–Fröhlich method. When dynamic corrections are incorporated into the computational scheme, a much better agreement with the experimental value of the dielectric constant is found when the corrected (real) dipole moment of liquid water is used. However, a formal analysis indicates that the static properties, such as dielectric constant, should not depend on dynamics. We discuss the resulting conundrum and related issues of simulations of electrostatic interactions using periodic boundary conditions in the context of our findings.

Cytochrome c oxidase (C cO) is the terminal enzyme in the respiratory electron transport chain. As part of its catalytic cycle, C cO transfers protons to its Fe-Cu binuclear center (BNC) to reduce oxygen, and in addition, it pumps protons across the mitochondrial inner, or bacterial, membrane where it is located. It is believed that this proton transport is facilitated by a network of water chains inside the enzyme. Here we present an analysis of the hydration of C cO, including the BNC region, using a semi-empirical hydration program, Dowser++, recently developed in our group. Using high-resolution X-ray data, we show that Dowser++ predictions match very accurately the water molecules seen in the D- and K-channels of C cO, as well as in the vicinity of its BNC. Moreover, Dowser++ predicts many more internal water molecules than is typically seen in the experiment. However, no significant hydration of the catalytic cavity in C cO described recently in the literature is observed. As Dowser++ itself does not account for structural changes of the protein, this result supports the earlier assessment that the proposed wetting transition in the catalytic cavity can only either be due to structural rearrangements of BNC, possibly induced by the charges during the catalytic cycle, or occur transiently, in concert with the proton transfer. Molecular dynamics simulations were performed to investigate the global dynamic nature of Dowser++ waters in C cO, and the results suggest a consistent explanation as to why some predicted water molecules would be missing in the experimental structures. Furthermore, in light of the significant protein hydration predicted by Dowser++, the dielectric constant of the hydrated cavities in C cO was also investigated using the Fröhlich-Kirkwood model; the results indicate that in the cavities where water is packed sufficiently densely the dielectric constant can approach values comparable even to that of bulk water.

No abstract