Self-Consistent Convolutional Density Functional Approximations: Application to Adsorption at Metal Surfaces.

The exchange-correlation (XC) functional in density functional theory is used to approximate multi-electron interactions. A plethora of different functionals are available, but nearly all are based on the hierarchy of inputs commonly referred to as "Jacob's ladder." This paper introduces an approach to construct XC functionals with inputs from convolutions of arbitrary kernels with the electron density, providing a route to move beyond Jacob's ladder. We derive the variational derivative of these functionals, showing consistency with the generalized gradient approximation (GGA), and provide equations for variational derivatives based on multipole features from convolutional kernels. A proof-of-concept functional, PBEq, which generalizes the PBE α ${\alpha }$ framework with α ${\alpha }$ being a spatially-resolved function of the monopole of the electron density, is presented and implemented. It allows a single functional to use different GGAs at different spatial points in a system, while obeying PBE constraints. Analysis of the results underlines the importance of error cancellation and the XC potential in data-driven functional design. After testing on small molecules, bulk metals, and surface catalysts, the results indicate that this approach is a promising route to simultaneously optimize multiple properties of interest.

A Universal Framework for Featurization of Atomistic Systems.

Machine-learning force fields have become increasingly popular because of their balance of accuracy and speed. However, a significant limitation is the use of element-specific features, leading to poor scalability with the number of elements. This work introduces the Gaussian multipole (GMP) featurization scheme that utilizes physically relevant multipole expansions of the electron density around atoms to yield feature vectors that interpolate between element types and have a fixed dimension regardless of the number of elements present. We combine GMP with neural networks and apply these models to the MD17 and QM9 data sets, revealing high computational efficiency, systematically improvable accuracy, and the ability to make reasonable predictions on elements not included in the training set. Finally, we test GMP-based models for the OCP data set, demonstrating comparable performance to graph-convolutional models. The results indicate that this featurization scheme fills a critical gap in the construction of efficient and transferable machine-learned force fields.

Novel inorganic crystal structures predicted using autonomous simulation agents.

We report a dataset of 96640 crystal structures discovered and computed using our previously published autonomous, density functional theory (DFT) based, active-learning workflow named CAMD (Computational Autonomy for Materials Discovery). Of these, 894 are within 1 meV/atom of the convex hull and 26826 are within 200 meV/atom of the convex hull. The dataset contains DFT-optimized pymatgen crystal structure objects, DFT-computed formation energies and phase stability calculations from the convex hull. It contains a variety of spacegroups and symmetries derived from crystal prototypes derived from known experimental compounds, and was generated from active learning campaigns of various chemical systems. This dataset can be used to benchmark future active-learning or generative efforts for structure prediction, to seed new efforts of experimental crystal structure discovery, or to construct new models of structure-property relationships.

Mapping the Pathway and Dynamics of Bestatin Inhibition of the Plasmodium falciparum M1 Aminopeptidase PfA-M1.

The M1 metallo-aminopeptidase from Plasmodium falciparum, PfA-M1, is an attractive drug target for the design of new antimalarials. Bestatin, a broad-spectrum metalloprotease inhibitor, is a moderate inhibitor of PfA-M1, and has been used to provide structure-activity relationships to inform drug design. The crystal structure of PfA-M1 with bestatin bound within its active site has been determined; however, dynamics of the inhibitor and the association or dissociation pathway have yet to be characterized. Here we present an all-atom molecular dynamics study where we have generated a hidden Markov state model from 2.3 µs of molecular dynamics simulation. Our hidden Markov state model identifies five macrostates that clearly show the events involved in bestatin dissociation from the PfA-M1 active site. The results show for the first time that bestatin can escape the substrate specificity pockets of the enzyme, primarily due to weak interactions within the pockets. Our approach identifies relevant conformational sampling of the inhibitor inside the enzyme and the protein dynamics that could be exploited to produce potent and selective inhibitors that can differentiate between similar members of the M1 aminopeptidase superfamily.

Generation of AMBER force field parameters for zinc centres of M1 and M17 family aminopeptidases.

The M1 and M17 aminopeptidases are metallo-exopeptidases that rely on the presence of divalent cations, usually zinc, in their active site for proteolytic activity. They are from separate protease superfamilies, however, members often have overlapping substrate specificity. Inhibitors of one or both enzymes can be used to modulate hypertension, reduce proliferation of certain types of cancers and control malaria parasites. Current inhibitors act to chelate the zinc ions in the active site, locking the enzymes in an inactive transition state. We were interested in using a computational approach to understand the structure and dynamics of the M1 and M17 aminopeptidases, however, the presence of the essential metal ions in the proteases presents a challenge to classical molecular dynamics (MD) simulation. The zinc amber force field does not contain applicable descriptions of the zinc coordination environment present in either of these two protease families. To provide tools for the study of these two enzymes, we have used the metal centre parameter builder to generate new hybrid bonded/nonbonded force field (FF) parameters to correctly describe the active site architecture for each enzyme. The new parameters were evaluated by fitting the normal mode frequencies of molecular mechanics to the quantum mechanics frequencies and validated by performing short MD simulations. The new FF parameters now enable more accurate and reliable MD simulations for any member of the M1 or M17 aminopeptidase superfamilies.

Mechanism of the Intramolecular Hexadehydro-Diels-Alder Reaction.

Theoretical analysis of the mechanism of the intramolecular hexadehydro-Diels-Alder (HDDA) reaction, validated against prior and newly measured kinetic data for a number of different tethered yne-diynes, indicates that the reaction proceeds in a highly asynchronous fashion. The rate-determining step is bond formation at the alkyne termini nearest the tether, which involves a transition-state structure exhibiting substantial diradical character. Whether the reaction then continues to close the remaining bond in a concerted fashion or in a stepwise fashion (i.e., with an intervening intermediate) depends on the substituents at the remaining terminal alkyne positions. Computational modeling of the HDDA reaction is complicated by the significant diradical character that arises along the reaction coordinate, which leads to instabilities in both restricted singlet Kohn-Sham density functional theory (DFT) and coupled cluster theory based on a Hartree-Fock reference wave function. A consistent picture emerges, however, from comparison of broken-symmetry DFT calculations and second-order perturbation theory based on complete-active-space self-consistent-field (CASPT2) calculations.

Efficient hydrolysis of corncob residue through cellulolytic enzymes from Trichoderma strain G26 and L-lactic acid preparation with the hydrolysate.

To prepare fermentable hydrolysate from corncob residue (CCR), Trichoderma strain G26 was cultured on medium containing CCR for production of cellulolytic enzymes through solid-state fermentation (SSF), resulting in 71.3 IU/g (FPA), 136.2 IU/g (CMCase), 85.1 IU/g (ß-glucosidase) and 11,344 IU/g (xylanase), respectively. Through a three-stage saccharification strategy, CCR was hydrolyzed by the enzymatic solution (6.5 FPU/ml) into fermentable hydrolysate containing 60.1g/l glucose (81.2% cellulose was converted at solid loading of 12.5%), 21.4% higher than that by the one-stage method. And then the hydrolysate was used to produce L-lactic acid by a previous screened strain Bacillus coagulans ZX25 in the submerged fermentation. 52.0 g/l L-lactic acid was obtained after fermentation for 44 h, with 86.5% glucose being converted to L-lactic acid. The results indicate that the strains and the hydrolysis strategy are promising for commercial production of L-lactic acid from CCR and other biomass.