Machine Learning to Predict Diels-Alder Reaction Barriers from the Reactant State Electron Density

Reaction barriers are key to our understanding of chemical reactivity and catalysis. Certain reactions are so seminal in chemistry that countless variants, with or without catalysts, have been studied, and their barriers have been computed or measured experimentally. This wealth of data represents a perfect opportunity to leverage machine learning models, which could quickly predict barriers without explicit calculations or measurement. Here, we show that the topological descriptors of the quantum mechanical charge density in the reactant state constitute a set that is both rigorous and continuous and can be used effectively for the prediction of reaction barrier energies to a high degree of accuracy. We demonstrate this on the Diels-Alder reaction, highly important in biology and medicinal chemistry, and as such, studied extensively. This reaction exhibits a range of barriers as large as 270 kJ/mol. While we trained our single-objective supervised (labeled) regression algorithms on simpler Diels-Alder reactions in solution, they predict reaction barriers also in significantly more complicated contexts, such a Diels-Alder reaction catalyzed by an artificial enzyme and its evolved variants, in agreement with experimental changes in k_cat. We expect this tool to apply broadly to a variety of reactions in solution or in the presence of a catalyst, for screening and circumventing heavily involved computations or experiments.

Electrostatic regulation of blue copper sites

In the last 50 years, the blue copper proteins became central targets of investigation. Extensive experiments focused on the Cu coordination to probe the effect of local perturbations on its properties. We found that local electric fields, generated by charged residues evolutionarily placed throughout the protein edifice, mainly second sphere, but also more remotely, constitute an additional significant factor regulating blue copper proteins. These fields are not random, but exhibit a highly specific directionality, negative with respect to the and vectors in the Cu first shell. The field magnitude contributes to fine-tuning of the geometric and electronic properties of Cu sites in individual blue copper proteins. Specifically, the local electric fields evidently control the Cu–S_Met bond distance, Cu(ii)–S_Cys bond covalency, and the energies of the frontier molecular orbitals, which, in turn, govern the Cu(ii/i) reduction potential and the relative absorption intensities at 450 nm and 600 nm.

Intramolecular electric fields in blue copper proteins are oriented in a fixed way to modulate properties of their copper sites: they control the first-shell copper interactions to influence geometric, spectroscopic, and redox behavior.

Contrasting Effects of Inhibitors Li+ and Be2+ on Catalytic Cycle of Glycogen Synthase Kinase-3β

Ionic lithium shows rare effectiveness for treating bipolar disorder and is a potential drug for neurodegenerative diseases. Unfortunately, lithium suffers from significant drawbacks, mainly a narrow therapeutic window. Among the targets of lithium, glycogen synthase kinase 3β (GSK-3β) may be responsible for its therapeutic effects. The development of alternative, selective inhibitors of this kinase could prevent lithium side effects, but such efforts have met little success so far. An atomistic understanding of Li⁺ inhibition and the GSK-3β phosphorylation reaction would therefore facilitate the development of new drugs. In this study, we use extensive sampling of catalytic states with our mixed quantum-classical dynamics method QM/DMD and binding affinities from a competitive metal affinity (CMA) approach to expand the atomistic picture of Li⁺ GSK-3β inhibition. We compare Li⁺ action with Be²⁺ and find our results in agreement with in vitro kinetics studies. Ultimately, our simulations show that Li⁺ inhibition is driven by decreasing the phosphorylation reaction rate, rather than reducing catalytic turnover through tight binding to different GSK-3β states like Be²⁺ inhibition. The effect of these metals derive from electrostatic differences and especially their smaller atomic radii compared to the native Mg²⁺ and thus provide insight for the development of GSK-3β inhibitors based on other paradigms.

Titr-DMD-A Rapid, Coarse-Grained Quasi-All-Atom Constant pH Molecular Dynamics Framework

The pH-dependence of enzyme fold stability and catalytic activity is a fundamentally dynamic, structural property which is difficult to study. The challenges and expense of investigating dynamic, atomic scale behavior experimentally means that computational methods, particularly constant pH molecular dynamics (CpHMD), are well situated tools for this. However, these methods often struggle with affordable sampling of sufficiently long time scales while also obtaining accurate pKa prediction and verifying the structures they generate. We introduce Titr-DMD, an affordable CpHMD method that combines the quasi-all-atom coarse-grained discrete molecular dynamics (DMD) method for conformational sampling with Propka for pKa prediction, to circumvent these issues. The combination enables rapid sampling on limited computational resources, while simulations are still performed on the atomic scale. We benchmark the method on a set of proteins with experimentally attested pKa and on the pH triggered conformational change in a staphylococcal nuclease mutant, a rare experimental study of such behavior. Our results show Titr-DMD to be an effective and inexpensive method to study pH-coupled protein dynamics.

Local Electric Fields as a Natural Switch of Heme-Iron Protein Reactivity

Heme-iron oxidoreductases operating through the high-valent Fe^IVO intermediates perform crucial and complicated transformations, such as oxidations of unreactive saturated hydrocarbons. These enzymes share the same Fe coordination, only differing by the axial ligation, e.g., Cys in P450 oxygenases, Tyr in catalases, and His in peroxidases. By examining ~200 heme-iron proteins, we show that the protein hosts exert highly specific intramolecular electric fields on the active sites, and there is a strong correlation between the direction and magnitude of this field and the protein function. In all heme proteins, the field is preferentially aligned with the Fe-O bond ( F_z ). The Cys-ligated P450 oxygenases have the highest average F_z of 28.5 MV cm^-1, i.e., most enhancing the oxyl-radical character of the oxo group, and consistent with the ability of these proteins to activate strong C-H bonds. In contrast, in Tyr-ligated proteins, the average F_z is only 3.0 MV cm^-1, apparently suppressing single-electron off-pathway oxidations, and in His-ligated proteins, F_z is -8.7 MV cm^-1. The operational field range is given by the trade-off between the low reactivity of the Fe^IVO Compound I at the more negative F_z , and the low selectivity at the more positive F_z . Consequently, a heme-iron site placed in the field characteristic of another heme-iron protein class loses its canonical function, and gains an adverse one. Thus, electric fields produced by the protein scaffolds, together with the nature of the axial ligand, control all heme-iron chemistry.

Visualizing the Dynamic Metalation State of New Delhi Metallo-β-lactamase-1 in Bacteria Using a Reversible Fluorescent Probe

New Delhi metallo-β-lactamase (NDM) grants resistance to a broad spectrum of β-lactam antibiotics, including last-resort carbapenems, and is emerging as a global antibiotic resistance threat. Limited zinc availability adversely impacts the ability of NDM-1 to provide resistance, but a number of clinical variants have emerged that are more resistant to zinc scarcity (e.g., NDM-15). To provide a novel tool to better study metal ion sequestration in host-pathogen interactions, we describe the development of a fluorescent probe that reports on the dynamic metalation state of NDM within Escherichia coli. The thiol-containing probe selectively coordinates the dizinc metal cluster of NDM and results in a 17-fold increase in fluorescence intensity. Reversible binding enables competition and time-dependent studies that reveal fluorescence changes used to detect enzyme localization, substrate and inhibitor engagement, and changes to metalation state through the imaging of live E. coli using confocal microscopy. NDM-1 is shown to be susceptible to demetalation by intracellular and extracellular metal chelators in a live-cell model of zinc dyshomeostasis, whereas the NDM-15 metalation state is shown to be more resistant to zinc flux. The development of this reversible turn-on fluorescent probe for the metalation state of NDM provides a new tool for monitoring the impact of metal ion sequestration by host defense mechanisms and for detecting inhibitor-target engagement during the development of therapeutics to counter this resistance determinant.

When Fluxionality Beats Size Selection: Acceleration of Ostwald Ripening of Sub-Nano Clusters

Size selection was demonstrated to suppress Ostwald ripening of supported catalytic nanoparticles. When the supported clusters are subnanometer in size and highly fluxional, such as Pt clusters on the rutile TiO₂ (110) surface, this paradigm breaks down, and the established theory of sintering needs a revision. At temperatures characteristic of catalysis (i.e. 700 K), sub-nano clusters thermally populate many low-energy metastable isomers. As these isomers all have different geometric and electronic structures, and thus, formation and dissociation energies (in lieu of surface energy), Ostwald ripening is not suppressed, despite the size-selection. However, some clusters arise as magic numbers in terms of sintering stability at the ensemble level. Acceleration of sintering by metastable species persists though weakens in polydisperse cluster systems. We propose a competing pathways theory for sintering, which at the atomistic level describes the found size-specific sintering behavior.

Regularization of least squares problems in CHARMM parameter optimization by truncated singular value decompositions

We examine the use of the truncated singular value decomposition and Tikhonov regularization in standard form to address ill-posed least squares problems Ax = b that frequently arise in molecular mechanics force field parameter optimization. We illustrate these approaches by applying them to dihedral parameter optimization of genotoxic polycyclic aromatic hydrocarbon-DNA adducts that are of interest in the study of chemical carcinogenesis. Utilizing the discrete Picard condition and/or a well-defined gap in the singular value spectrum when A has a well-determined numerical rank, we are able to systematically determine truncation and in turn regularization parameters that are correspondingly used to produce truncated and regularized solutions to the ill-posed least squares problem at hand. These solutions in turn result in optimized force field dihedral terms that accurately parameterize the torsional energy landscape. As the solutions produced by this approach are unique, it has the advantage of avoiding the multiple iterations and guess and check work often required to optimize molecular mechanics force field parameters.

Optical Cycling Functionalization of Arenes

Closed, laser-induced optical transitions ("optical cycling transitions") of molecules can be used for state preparation and measurement in quantum information science and quantum sensing. Increasingly complex molecular species supporting optical cycling can provide new capabilities for quantum science, and it is not clear if there is a limit on their size or complexity. We explore Ca-O-L molecular constructs to support the optical cycling center, Ca, with ligands, L, being arenes. We find that L can be as large as coronene (i.e., CaOC₂₄H₁₁) without losing the diagonality of the Franck-Condon factor (FCF). Furthermore, L can be substituted with electron-withdrawing groups to improve the FCF. Larger L, beyond ∼7 rings, can disrupt the diagonality of the FCF by closing the HOMO-LUMO ligand electronic state gap and reordering with the local states on the cycling center. Overall, we find that optical cycling can be retained for arenes, and we offer a principle for their design.

Franck-Condon Tuning of Optical Cycling Centers by Organic Functionalization

Laser induced electronic excitations that spontaneously emit photons and decay directly to the initial ground state ("optical cycling transitions") are used in quantum information and precision measurement for state initialization and readout. To extend this primarily atomic technique to large, organic compounds, we theoretically investigate optical cycling of alkaline earth phenoxides and their functionalized derivatives. We find that optical cycle leakage due to wave function mismatch is low in these species, and can be further suppressed by using chemical substitution to boost the electron-withdrawing strength of the aromatic molecular ligand through resonance and induction effects. This provides a straightforward way to use chemical functional groups to construct optical cycling moieties for laser cooling, state preparation, and quantum measurement.

Surface chemical trapping of optical cycling centers

Quantum information processors based on trapped atoms utilize laser-induced optical cycling transitions for state preparation and measurement. These transitions consist of an electronic excitation from the ground to an excited state and a decay back to the initial ground state, associated with a photon emission. While this technique has been used primarily with atoms, it has also recently been shown to work for some divalent metal hydroxides (e.g. SrOH) and alkoxides (e.g. SrOCH3). This extension to molecules is possible because these molecules feature nearly isolated, atomic-like ground and first-excited electronic states centered on the radical metal atom. We theoretically investigate the extension of this idea to a larger scale by growing the alkyl group, R, beyond the initial methyl group, CH3, while preserving the isolated and highly vertical character of the electronic excitation on the radical metal atom, M. Theory suggests that in the limit as the size of the ligand carbon chain increases, it can be considered a functionalized diamond (or cubic boron nitride) surface. Several requirements must be observed for the cycling centers to function when bound to the surface. First, the surface must have a significant band gap that fully encapsulates both the ground and excited states of the cycling center. Second, while the surface lattice imposes strict limits on the achievable spacing between the SrO- groups, at high coverage, SrO- centers can interact, and show geometric changes and/or electronic state mixing. We show that the coverage of the diamond surface with SrO- cycling centers needs to be significantly sub-monolayer for the functionality of the cycling center to be preserved. Having the lattice-imposed spatial control of SrO- placements will allow nanometer-scale proximity between qubits and will eliminate the need for atom traps for localized cycling emitters. Our results also imply that a functionalization could be done on a scanning microscope tip for local quantum sensing or on photonic structures for optically-mediated quantum information processing.