Data-Efficient, Chemistry-Aware Machine Learning Predictions of Diels-Alder Reaction Outcomes.

The application of machine learning models to the prediction of reaction outcomes currently needs large and/or highly featurized data sets. We show that a chemistry-aware model, NERF, which mimics the bonding changes that occur during reactions, allows for highly accurate predictions of the outcomes of Diels-Alder reactions using a relatively small training set, with no pretraining and no additional features. We establish a diverse data set of 9537 intramolecular, hetero-, aromatic, and inverse electron demand Diels-Alder reactions. This data set is used to train a NERF model, and the performance is compared against state-of-the-art classification and generative machine learning models across low- and high-data regimes, with and without pretraining. The predictive accuracy (regio- and site selectivity in the major product) achieved by NERF exceeds 90% when as little as 40% of the data set is used for training. Another high-performing model, Chemformer, requires a larger training data set (>45%) and pretraining to reach 90% Top-1 accuracy. Accurate predictions of less-represented reaction subclasses, such as those involving heteroatomic or aromatic substrates, require higher percentages of training data. We also show how NERF can use small amounts of additional training data to quickly learn new systems and improve its overall understanding of reactivity. Synthetic chemists stand to benefit as this model can be rapidly expanded and tailored to areas of chemistry corresponding to the low-data regime.

TS-tools: Rapid and automated localization of transition states based on a textual reaction SMILES input.

Here, TS-tools is presented, a Python package facilitating the automated localization of transition states (TS) based on a textual reaction SMILES input. TS searches can either be performed at xTB or DFT level of theory, with the former yielding guesses at marginal computational cost, and the latter directly yielding accurate structures at greater expense. On a benchmarking dataset of mono- and bimolecular reactions, TS-tools reaches an excellent success rate of 95% already at xTB level of theory. For tri- and multimolecular reaction pathways - which are typically not benchmarked when developing new automated TS search approaches, yet are relevant for various types of reactivity, cf. solvent- and autocatalysis and enzymatic reactivity - TS-tools retains its ability to identify TS geometries, though a DFT treatment becomes essential in many cases. Throughout the presented applications, a particular emphasis is placed on solvation-induced mechanistic changes, another issue that received limited attention in the automated TS search literature so far.

Machine Learning-Guided Computational Screening of New Candidate Reactions with High Bioorthogonal Click Potential.

Bioorthogonal click chemistry has become an indispensable part of the biochemist's toolbox. Despite the wide variety of applications that have been developed in recent years, only a limited number of bioorthogonal click reactions have been discovered so far, most of them based on (substituted) azides. In this work, we present a computational workflow to discover new candidate reactions with promising kinetic and thermodynamic properties for bioorthogonal click applications. Sampling only around 0.05 % of an overall search space of over 10,000,000 dipolar cycloadditions, we develop a machine learning model able to predict DFT-computed activation and reaction energies within ∼2-3 kcal/mol across the entire space. Applying this model to screen the full search space through iterative rounds of learning, we identify a broad pool of candidate reactions with rich structural diversity, which can be used as a starting point or source of inspiration for future experimental development of both azide-based and non-azide-based bioorthogonal click reactions.

Combining Molecular Quantum Mechanical Modeling and Machine Learning for Accelerated Reaction Screening and Discovery.

Molecular quantum mechanical modeling, accelerated by machine learning, has opened the door to high-throughput screening campaigns of complex properties, such as the activation energies of chemical reactions and absorption/emission spectra of materials and molecules; in silico. Here, we present an overview of the main principles, concepts, and design considerations involved in such hybrid computational quantum chemistry/machine learning screening workflows, with a special emphasis on some recent examples of their successful application. We end with a brief outlook of further advances that will benefit the field.

Quantum chemistry-augmented neural networks for reactivity prediction: Performance, generalizability, and explainability.

There is a perceived dichotomy between structure-based and descriptor-based molecular representations used for predictive chemistry tasks. Here, we study the performance, generalizability, and explainability of the quantum mechanics-augmented graph neural network (ml-QM-GNN) architecture as applied to the prediction of regioselectivity (classification) and of activation energies (regression). In our hybrid QM-augmented model architecture, structure-based representations are first used to predict a set of atom- and bond-level reactivity descriptors derived from density functional theory calculations. These estimated reactivity descriptors are combined with the original structure-based representation to make the final reactivity prediction. We demonstrate that our model architecture leads to significant improvements over structure-based GNNs in not only overall accuracy but also in generalization to unseen compounds. Even when provided training sets of only a couple hundred labeled data points, the ml-QM-GNN outperforms other state-of-the-art structure-based architectures that have been applied to these tasks as well as descriptor-based (linear) regressions. As a primary contribution of this work, we demonstrate a bridge between data-driven predictions and conceptual frameworks commonly used to gain qualitative insights into reactivity phenomena, taking advantage of the fact that our models are grounded in (but not restricted to) QM descriptors. This effort results in a productive synergy between theory and data science, wherein QM-augmented models provide a data-driven confirmation of previous qualitative analyses, and these analyses in turn facilitate insights into the decision-making process occurring within ml-QM-GNNs.

Promotion Energy Analysis Predicts Reaction Modes: Nucleophilic and Electrophilic Aromatic Substitution Reactions.

To develop an approach to pre-emptively predict the existence of major reaction modes associated with a chemical system, based on exclusive consideration of reactant properties, we build herein on the valence bond perspective of chemical reactivity. In this perspective, elementary chemical reactions are conceptualized as crossovers between individual diabatic/semilocalized states. As demonstrated, the spacings between the main diabatic states in the reactant geometries-the so-called promotion energies-contain predictive information about which types of crossings are likely to occur on a potential energy surface, facilitating the identification of potential transition states and products. As an added bonus, promotion energy analysis provides direct insight into the impact of environmental effects, e.g., the presence of (polar) solvents and/or (local) electric fields, on a mechanistic landscape. We illustrate the usefulness of our approach by focusing on model nucleophilic and electrophilic aromatic substitution reactions. Overall, we envision our analysis to be useful not only as a tool for conceptualizing individual mechanistic landscapes but also as a facilitator of systematic reaction-network exploration efforts. Because the emerging VB descriptors are computationally inexpensive (and can alternatively be inferred through machine learning), they could be evaluated on-the-fly as part of an exploration algorithm. The so-predicted reaction modes could subsequently be examined in detail through computationally more-demanding methods.

Resolving Entangled Reactivity Modes through External Electric Fields and Substitution: Application to E2/SN2 Reactions.

In this study, we explore strategies to resolve entangled reactivity modes. More specifically, we consider the competition between SN2 and E2 reaction pathways for alkyl halides and nucleophiles/bases. We first demonstrate that the emergence of an E2-preference is associated with an enhancement of the magnitude of the resonance stabilization in the transition-state (TS) region, resulting from the improved mixing of electrostatically stabilized valence bond structures into the TS wavefunction. Subsequently, we show that the TS resonance energy can be tuned selectively and rationally either through the application of an oriented external electric field directed along the C-C axis of the alkyl halide or through a regular substitution approach of the C-C moiety. We end our study by demonstrating that the insights gained from our analysis enable one to rationalize the main reactivity trends emerging from a recently published large database of competing SN2 and E2 reaction pathways.

Do Diradicals Behave Like Radicals?

This review sets out to understand the reactivity of diradicals and how that may differ from monoradicals. In the first part of the review, we delineate the electronic structure of a diradical with its two degenerate or nearly degenerate molecular orbitals, occupied by two electrons. A classification of diradicals based on whether or not the two SOMOs can be located on different sites of the molecule is useful in determining the ground state spin. Important is a delocalized to localized orbital transformation that interchanges "closed-shell" to "open-shell" descriptions. The resulting duality is useful in understanding the dual reactivity of singlet diradicals. In the second part of the review, we examine, with a consistent level of theory, activation energies of prototypical radical reactions (dimerization, hydrogen abstraction, and addition to ethylene) for representative organic diradicals and diradicaloids in their two lowest spin states. Differences and similarities in reactivity of diradicals vs monoradicals, based on either a localized or delocalized view, whichever is suitable, are then discussed. The last part of this review begins with an extensive, comparative, and critical survey of available measures of diradical character and ends with an analysis of the consequences of diradical character for selected diradicaloids.

Extending conceptual DFT to include additional variables: oriented external electric field.

The extension of the E = E[N, v] functional for exploring chemical reactivity in a conceptual DFT context to include external electric fields is discussed. Concentrating on the case of a homogeneous field the corresponding response functions are identified and integrated, together with the conventional response functions such as permanent dipole moment and polarizability, in an extended response function tree associated with the E = E[N, v, ε] functional. In a case study on the dihalogens F2, Cl2, Br2, I2 the sensitivity of condensed atomic charges (∂q/∂ε) is linked to the polarizability of the halogen atoms. The non-integrated (∂ρ(r)/∂ε) response function, directly related to the field induced density change, is at the basis of these features. It reveals symmetry breaking for a perpendicular field, not detectable in its atom condensed counterpart, and accounts for the induced dipole moment directly related to the molecular polarizability. The much higher sensitivity of the electronic chemical potential/electronegativity as compared to the chemical hardness is highlighted. The response of the condensed Fukui functions to a parallel electric field increases when going down in the periodic table and is interpreted in terms of the extension of the outer contours in the non-condensed Fukui function. In the case of a perpendicular field the (∂f(r)/∂ε) response function hints at stereoselectivity with a preferential side of attack which is not retrieved in its condensed form. In an application the nucleophilic attack on the carbonyl group in H2CO is discussed. Similar to the dihalogens, stereoselectivity is displayed in the Fukui function for nucleophilic attack (f+) in the case of a perpendicular electric field, and opposite to the one that would arise based on the induced density. Disentangling the expression for the evolution of the Fukui function in the presence of an electric field reveals that this difference can be traced back to local differences in the polarization or induced density between the anionic and the neutral system. This difference may be exploited, e.g. for an appropriately substituted H2CO, to generate enantioselectivity.

Unifying Conceptual Density Functional and Valence Bond Theory: The Hardness-Softness Conundrum Associated with Protonation Reactions and Uncovering Complementary Reactivity Modes.

In this study, we address the long-standing issue-arising prominently from conceptual density functional theory (CDFT)-of the relative importance of electrostatic, i.e., "hard-hard", versus spin-pairing, i.e., "soft-soft", interactions in determining regiochemical preferences. We do so from a valence bond (VB) perspective and demonstrate that VB theory readily enables a clear-cut resolution of both of these contributions to the bond formation/breaking process. Our calculations indicate that appropriate local reactivity descriptors can be used to gauge the magnitude of both interactions individually, e.g., Fukui functions or HOMO/LUMO orbitals for the spin-pairing/(frontier) orbital interactions and molecular electrostatic potentials (and/or partial charges) for the electrostatic interactions. In contrast to previous reports, we find that protonation reactions cannot generally be classified as either charge- or frontier orbital-controlled; instead, our results indicate that these two bonding contributions generally interplay in more subtle patterns, only giving the impression of a clear-cut dichotomy. Finally, we demonstrate that important covalent, i.e., spin pairing, reactivity modes can be missed when only a single spin-pairing/orbital interaction descriptor is considered. This study constitutes an important step in the unification of CDFT and VB theory.

Oriented External Electric Fields and Ionic Additives Elicit Catalysis and Mechanistic Crossover in Oxidative Addition Reactions.

Judiciously applied oriented external electric fields (OEEFs) exert catalytic effects on the kinetics and improve the thermodynamics of chemical reactions. Herein, we examine the ability of OEEFs to assist catalysts and show that the rate of oxidative addition between palladium catalysts and alkyl/aryl electrophiles can be controlled by an OEEF applied along the direction of electron reorganization (the "reaction axis"). The concerted mechanism of oxidative addition proceeds through a transition state with moderate charge transfer character. We demonstrate that OEEFs along the reaction axis can control this charge transfer and impart electrostatic catalysis. When the applied field exceeds a certain critical value (∼0.15 V/Å), we observed a mechanistic crossover from the concerted to a dissociative CSNAr type of reactivity for aryl electrophiles. To our surprise, alkyl electrophiles follow a hitherto unexplored SN2 pathway for the reaction with large transition state stabilization at relatively low OEEFs. A valence-bond state correlation diagram (VBSCD) is employed to comprehend the results. Finally, although the catalytic effect of salt additives in oxidative addition is known, its mechanism is still under debate. Our findings further show evidence that salt additives exert electric-field effects on the rate of cross-coupling reactions, and their cocatalytic effects can be judiciously reproduced by applied external electric fields. As such, we propose that the use of additives (anionic or cationic) is an experimentally viable strategy to implement external electric-field effects in routinely used oxidative addition catalysis.

Electric-Field Mediated Chemistry: Uncovering and Exploiting the Potential of (Oriented) Electric Fields to Exert Chemical Catalysis and Reaction Control.

This Perspective discusses oriented external-electric-fields (OEEF), and other electric-field types, as "smart reagents", which enable in principle control over wide-ranging aspects of reactivity and structure. We discuss the potential of OEEFs to control nonredox reactions and impart rate-enhancement and selectivity. An OEEF along the "reaction axis", which is the direction whereby electronic reorganization converts reactants' to products' bonding, will accelerate reactions, control regioselectivity, induce spin-state selectivity, and elicit mechanistic crossovers. Simply flipping the direction of the OEEF will lead to inhibition. Orienting the OEEF off the reaction axis enables control over stereoselectivity, enantioselectivity, and product selectivity. For polar/polarizable reactants, the OEEF itself will act as tweezers, which orient the reactants and drive their reaction. OEEFs also affect bond-dissociation energies and dissociation modes (covalent vs ionic), as well as alteration of molecular geometries and supramolecular aggregation. The "key" to gaining access to this toolbox provided by OEEFs is microscopic control over the alignment between the molecule and the applied field. We discuss the elegant experimental methods which have been used to verify the theoretical predictions and describe various alternative EEF sources and prospects for upscaling OEEF catalysis in solvents. We also demonstrate the numerous ways in which the OEEF effects can be mimicked by use of (designed) local-electric fields (LEFs), i.e., by embedding charges or dipoles into molecules. LEFs and OEEFs are shown to be equivalent and to obey the same ground rules. Outcomes are exemplified for Diels-Alder cycloadditions, oxidative addition of bonds by transition-metal complexes, H-abstractions by oxo-metal species, ionic cleavage of halogen bonds, methane activation, etc.

How Do Local Reactivity Descriptors Shape the Potential Energy Surface Associated with Chemical Reactions? The Valence Bond Delocalization Perspective.

How do local reactivity descriptors, such as the Fukui function and the local spin density distribution, shape the potential energy surface (PES) associated with chemical reactions and thus govern reactivity trends and regioselective preferences? This is the question that is addressed here through a qualitative valence bond (VB) analysis. We demonstrate that common density functional theory (DFT)-based local reactivity descriptors can essentially be regarded-in one way or another-as indirect measures of delocalization, i.e., resonance stabilization, of the reactants within VB theory. The inherent connection between (spatial) delocalization and (energetic) resonance stabilization embedded in VB theory provides a natural and elegant framework for analyzing and comprehending the impact of individual local reactivity descriptors on the global PES. Our analysis provides new insights into the role played by local reactivity descriptors and illustrates under which conditions they can sometimes fail to predict reactivity trends and regioselective preferences, e.g., in the case of ambident reactivity. This treatment constitutes a first step toward a unification of VB theory and conceptual DFT.

Solvent Organization and Rate Regulation of a Menshutkin Reaction by Oriented External Electric Fields are Revealed by Combined MD and QM/MM Calculations.

When and how do external electric fields (EEFs) lead to catalysis in the presence of a (polar or nonpolar) solvent? This is the question that is addressed here using a combination of molecular dynamics (MD) simulations, quantum mechanical/molecular mechanical calculations with EEF, and quantum mechanical/(local) electric field calculations. The paper focuses on a model reaction, the Menshutkin reaction between CH3I and pyridine in three solvents of varying polarity. Using MD simulations, we find that the EEF causes the solvent to undergo organization; the solvent molecules gradually align with the applied field as the field strength increases. The collective orientation of the solvent molecules modifies the electrostatic environment around the Menshutkin species and induces a global electric field pointing in the opposite direction of the applied EEF. The combination of these two entangled effects leads to partial or complete screening of the EEF, with the extent of screening being proportional to the polarity/polarizability of the solvent. Nevertheless, we find that catalysis of the Menshutkin reaction inevitably emerges once the EEF exceeds the opposing field of the organizing solvent, i.e., once polarization of the Menshutkin complex is observed to set in. Overall, our analysis provides a lucid and pictorial interpretation of the behavior of solutions in the presence of EEFs and indicates that EEF-mediated catalysis should, in principle, be feasible in bulk setups, especially for nonpolar and mildly polar solvents. By application of the charge-transfer paradigm, it is shown that the emergence of OEEF catalysis in solution can be generalized to other reactions as well.

TITAN: A Code for Modeling and Generating Electric Fields-Features and Applications to Enzymatic Reactivity.

We present here a versatile computational code named "elecTric fIeld generaTion And maNipulation (TITAN)," capable of generating various types of external electric fields, as well as quantifying the local (or intrinsic) electric fields present in proteins and other biological systems according to Coulomb's Law. The generated electric fields can be coupled with quantum mechanics (QM), molecular mechanics (MM), QM/MM, and molecular dynamics calculations in most available software packages. The capabilities of the TITAN code are illustrated throughout the text with the help of examples. We end by presenting an application, in which the effects of the local electric field on the hydrogen transfer reaction in cytochrome P450 OleTJE enzyme and the modifications induced by the application of an oriented external electric field are examined. We find that the protein matrix in P450 OleTJE acts as a moderate catalyst and that orienting an external electric field along the Fe─O bond of compound I has the biggest impact on the reaction barrier. The induced catalysis/inhibition correlates with the calculated spin density on the O-atom. © 2019 Wiley Periodicals, Inc.

Oriented (Local) Electric Fields Drive the Millionfold Enhancement of the H-Abstraction Catalysis Observed for Synthetic Metalloenzyme Analogues.

This contribution follows the recent remarkable catalysis observed by Groves et al. in hydrogen-abstraction reactions by a) an oxoferryl porphyrin radical-cation complex [Por⋅+ FeIV (O)Lax ] and b) a hydroxoiron porphyrazine ferric complex [PyPzFeIII (OH)Lax ], both of which involve positively charged substituents on the outer circumference of the respective macrocyclic ligands. These charge-coronated complexes are analogues of the biologically important Compound I (Cpd I) and synthetic hydroxoferric species, respectively. We demonstrate that the observed enhancement of the H-abstraction catalysis for these systems is a purely electrostatic effect, elicited by the local charges embedded on the peripheries of the respective macrocyclic ligands. Our findings provide new insights into how electrostatics can be employed to tune the catalytic activity of metalloenzymes and can thus contribute to the future design of new and highly efficient hydrogen-abstraction catalysts.

Electrophilic Aromatic Substitution Reactions: Mechanistic Landscape, Electrostatic and Electric-Field Control of Reaction Rates, and Mechanistic Crossovers.

This study investigates the rich mechanistic landscape of the iconic electrophilic aromatic substitution (EAS) reaction class, in the gas phase, in solvents, and under stimulation by oriented external electric fields. The study uses DFT calculations, complemented by a qualitative valence bond (VB) perspective. We construct a comprehensive and unifying framework that elucidates the many surprising mechanistic features, uncovered in recent years, of this class of reactions. For example, one of the puzzling issues which have attracted significant interest recently is the finding of a variety of concerted mechanisms that do not involve the formation of σ-complex intermediates, in apparent contradiction to the generally accepted textbook mechanism. Our VB modeling elucidates the existence of both the concerted and stepwise mechanisms and uncovers the root causes and necessary conditions for the appearance of these intermediates. Furthermore, our VB analysis offers insight into the potential applications of external electric fields as smart, green, and selective catalysts, which can control at will reaction rates, as well as mechanistic crossovers, for this class of reactions. Finally, we highlight how understanding of the electric fields effect on the EAS reaction could lead to the formulation of guiding principles for the design of improved heterogeneous catalysts. Overall, our analysis underscores the powerful synergy offered by combining molecular orbital and VB theory to tackle interesting and challenging mechanistic questions in chemistry.

Cross Conjugation in Polyenes and Related Hydrocarbons: What Can Be Learned from Valence Bond Theory about Single-Molecule Conductance?

This study examined the nature of the electronic structure of representative cross-conjugated polyenes from a valence bond (VB) perspective. Our VBSCF calculations on a prototypical dendralene model reveal a remarkable inhibition of the delocalization compared to linear polyenes. Especially along the C-C backbone, the delocalization is virtually quenched so that these compounds can essentially be considered as sets of isolated butadiene units. In direct contrast to the dendralene chains, quinodimethane compounds exhibit an enhancement in their delocalization compared to linear polyenes. We demonstrate that this quenching/enhancement of the delocalization is inherently connected to the relative weights of specific types of long-bond VB structures. From our ab initio treatment, many localization/delocalization-related concepts and phenomena, central to both organic chemistry and single-molecule electronics, emerge. Not only do we find direct insight into the relation between topology and the occurrence of quantum interference (QI), but we also find a phenomenological justification of the recently proposed diradical character-based rule for the estimation of the magnitude of molecular conductance. Generally, our results can be conceptualized using the "arrow-pushing" concept, originating from resonance theory.

Global and local aromaticity of acenes from the information-theoretic approach in density functional reactivity theory.

In this work, we report a systematic study on the global and local aromaticity of acenes using a series of model structures from 2-acene to 11-acene. A recently developed ansatz, an information-theoretic approach coached into density functional reactivity theory has been employed, which essentially provides different density functionals characterizing the molecular electron density distribution. Based on the correlation analysis of six conventional aromaticity indices with eight information-theoretic quantities, we examined the aromaticity of acenes from both global and local perspectives. From the global aromaticity viewpoint, our results suggest that different descriptors based on various physicochemical properties are intrinsically dependent. A novel laminated feature ruling local aromaticity of acenes has been unveiled, from which we found that the distance from the terminal rings plays the critical role. Based on the shape of the correlation plots between the conventional aromaticity indices and information-theoretic quantities, the latter could be separated into three subgroups. The seemingly contradictory results from global and local aromaticity perspectives not only present us the uniqueness of the acene systems but all demonstrate the effectiveness of the information-theoretic approach from density functional reactivity theory. Besides strengthening the validity of a series of new aromaticity descriptors, our results should lead to more clear insights into the chemical significance of the information-theoretic quantities.

Captodative Substitution Enhances the Diradical Character of Compounds, Reduces Aromaticity, and Controls Single-Molecule Conductivity Patterns: A Valence Bond Study.

The present contribution uses a valence bond (VB) perspective to consider the captodative substitution strategy, a method to enhance the diradical character of (potentially aromatic) compounds. We confirm the qualitative reasoning that has generally been used to rationalize the diradical-character-enhancing effect of captodative substitution: this type of substitution scheme disproportionally stabilizes specific Dewar/diradical(oid) VB structures, thus increasing their weight in the full ground-state wave function. Furthermore, we assess the effect of captodative substitution on the aromaticity of the considered compound. We observe a clear trade-off between diradical character and aromaticity for our model systems: as one of these properties increases, the other decreases. This finding is especially significant within the field of single-molecule electronics because it enables unification of the previously observed inverse proportionality between the aromaticity of a compound and the magnitude of conductance through that molecule, with the observed proportionality between diradical character and the magnitude of conductance associated with a compound. To some extent, both properties, i.e., aromaticity and diradical character, appear to be the flip-sides of the same coin.