Computational Study of the Ir-Catalyzed Formation of Allyl Carbamates from CO2.

We have employed computational methods to investigate the iridium-catalyzed allylic substitution leading to the formation of enantioenriched allyl carbamates from carbon dioxide (CO2). The reaction occurs in several steps, with initial formation of an iridium-allyl, followed by nucleophilic attack by the carbamate formed in situ from CO2 and an amine. A detailed isomeric analysis shows that the rate-determining step differs for the (R)- and (S)-pathways. These insights are essential for understanding reactions involving enantioselective formation of allyl carbamates from CO2.

Nickel Catalyzed Carbonylative Cross Coupling for Direct Access to Isotopically Labeled Alkyl Aryl Ketones.

Here we present an effective nickel-catalyzed carbonylative cross-coupling for direct access to alkyl aryl ketones from readily accessible redox-activated tetrachlorophthalimide esters and aryl boronic acids. The methodology, which is run employing only 2.5 equivalents of CO and simple Ni(II) salts as the metal source, exhibits a broad substrate scope under mild conditions. Furthermore, this carbonylation chemistry provides an easy switch between isotopologues for stable (13CO) and radioactive (14CO) isotope labeling, allowing its adaptation to the late-stage isotope labeling of pharmaceutically relevant compounds. Based on DFT calculations as well as experimental evidence, a catalytic cycle is proposed involving a carbon-centered radical formed via nickel(I)-induced outer-sphere decarboxylative fragmentation of the redox-active ester.

Finding Relevant Retrosynthetic Disconnections for Stereocontrolled Reactions.

Machine learning-driven computer-aided synthesis planning (CASP) tools have become important tools for idea generation in the design of complex molecule synthesis but do not adequately address the stereochemical features of the target compounds. A novel approach to automated extraction of templates used in CASP that includes stereochemical information included in the US Patent and Trademark Office (USPTO) and an internal AstraZeneca database containing reactions from Reaxys, Pistachio, and AstraZeneca electronic lab notebooks is implemented in the freely available AiZynthFinder software. Three hundred sixty-seven templates covering reagent- and substrate-controlled as well as stereospecific reactions were extracted from the USPTO, while 20,724 templates were from the AstraZeneca database. The performance of these templates in multistep CASP is evaluated for 936 targets from the ChEMBL database and an in-house selection of 791 AZ designs. The potential and limitations are discussed for four case studies from ChEMBL and examples of FDA-approved drugs.

Kinetically-Controlled Ni-Catalyzed Direct Carboxylation of Unactivated Secondary Alkyl Bromides without Chain Walking.

Herein, we report the direct carboxylation of unactivated secondary alkyl bromides enabled by the merger of photoredox and nickel catalysis, a previously inaccessible endeavor in the carboxylation arena. Site-selectivity is dictated by a kinetically controlled insertion of CO2 at the initial C(sp3)-Br site by the rapid formation of Ni(I)-alkyl species, thus avoiding undesired ß-hydride elimination and chain-walking processes. Preliminary mechanistic experiments reveal the subtleties of stereoelectronic effects for guiding the reactivity and site-selectivity.

Regioselective ortho halogenation of N-aryl amides and ureas via oxidative halodeboronation: harnessing boron reactivity for efficient C-halogen bond installation.

The installation of the C-halogen bond at the ortho position of N-aryl amides and ureas represents a tool to prepare motifs that are ubiquitous in biologically active compounds. To construct such prevalent bonds, most methods require the use of precious metals and a multistep process. Here we report a novel protocol for the long-standing challenge of regioselective ortho halogenation of N-aryl amides and ureas using an oxidative halodeboronation. By harnessing the reactivity of boron over nitrogen, we merge carbonyl-directed borylation with consecutive halodeboronation, enabling the precise introduction of the C-X bond at the desired ortho position of N-aryl amides and ureas. This method offers an efficient, practical, and scalable solution for synthesizing halogenated N-heteroarenes under mild conditions, highlighting the superiority of boron reactivity in directing the regioselectivity of the reaction.

Hybrid Machine Learning Approach to Predict the Site Selectivity of Iridium-Catalyzed Arene Borylation.

The borylation of aryl and heteroaryl C-H bonds is valuable for the site-selective functionalization of C-H bonds in complex molecules. Iridium catalysts ligated by bipyridine ligands catalyze the borylation of the C-H bond that is most acidic and least sterically hindered in an arene, but predicting the site of borylation in molecules containing multiple arenes is difficult. To address this challenge, we report a hybrid computational model that predicts the Site of Borylation (SoBo) in complex molecules. The SoBo model combines density functional theory, semiempirical quantum mechanics, cheminformatics, linear regression, and machine learning to predict site selectivity and to extrapolate these predictions to new chemical space. Experimental validation of SoBo showed that the model predicts the major site of borylation of pharmaceutical intermediates with higher accuracy than prior machine-learning models or human experts, demonstrating that SoBo will be useful to guide experiments for the borylation of specific C(sp2)-H bonds during pharmaceutical development.

On the use of real-world datasets for reaction yield prediction.

The lack of publicly available, large, and unbiased datasets is a key bottleneck for the application of machine learning (ML) methods in synthetic chemistry. Data from electronic laboratory notebooks (ELNs) could provide less biased, large datasets, but no such datasets have been made publicly available. The first real-world dataset from the ELNs of a large pharmaceutical company is disclosed and its relationship to high-throughput experimentation (HTE) datasets is described. For chemical yield predictions, a key task in chemical synthesis, an attributed graph neural network (AGNN) performs as well as or better than the best previous models on two HTE datasets for the Suzuki-Miyaura and Buchwald-Hartwig reactions. However, training the AGNN on an ELN dataset does not lead to a predictive model. The implications of using ELN data for training ML-based models are discussed in the context of yield predictions.

AiZynthTrain: Robust, Reproducible, and Extensible Pipelines for Training Synthesis Prediction Models.

We introduce the AiZynthTrain Python package for training synthesis models in a robust, reproducible, and extensible way. It contains two pipelines that create a template-based one-step retrosynthesis model and a RingBreaker model that can be straightforwardly integrated in retrosynthesis software. We train such models on the publicly available reaction data set from the U.S. Patent and Trademark Office (USPTO), and these are the first retrosynthesis models created in a completely reproducible end-to-end fashion, starting with the original reaction data source and ending with trained machine-learning models. In particular, we show that employing new heuristics implemented in the pipeline greatly improves the ability of the RingBreaker model for disconnecting ring systems. Furthermore, we demonstrate the robustness of the pipeline by training on a more diverse but proprietary data set. We envisage that this framework will be extended with other synthesis models in the future.

A Quantum-Guided Molecular Mechanics Force Field for the Ferrocene Scaffold.

Ferrocene derivatives have a wide range of applications, including as ligands in asymmetric catalysis, due to their chemical stability, rigid backbone, steric bulk, and ability to encode stereochemical information via planar chirality. Unfortunately, few of the available molecular mechanics force fields incorporate parameters for the accurate study of this important building block. Here, we present a MM3* force field for ferrocenyl ligands, which was generated using the quantum-guided molecular mechanics (Q2MM) method. Detailed validation by comparison to DFT calculations and crystal structures demonstrates the accuracy of the parameters and uncovers the physical origin of deviations through excess energy analysis. Combining the ferrocene force field with a force field for Pd-allyl complexes and comparing the crystal structures shows the compatibility with previously developed MM3* force fields. Finally, the ferrocene force field was combined with a previously published transition-state force field to predict the stereochemical outcomes of the aminations of Pd-allyl complexes with different amines and different chiral ferrocenyl ligands, with an R2 of ∼0.91 over 10 examples.

Explaining Regiodivergent Vinylations with Vinylbenziodoxolones.

Vinylbenziodoxolones have recently been identified as efficient hypervalent iodine(III) reagents for electrophilic vinylations under transition metal-free conditions. Their unique reactivity allows synthesis of either internal or terminal alkenes, depending on the nucleophile class. This paper constitutes the first mechanistic investigation of VBX vinylations, and makes use of NMR studies, deuterium labelling and computations to rationalize the observed regio- and stereochemical outcome. Internal alkene formation in S-vinylation was found to proceed through the ligand coupling mechanism typical of diaryliodonium salts, whereas terminal alkene formation in P-vinylations took place via a phosphinous acid-coordinated VBX complex, which underwent concerted deprotonation and Michael-type addition. Subsequent base-assisted protonation and E2 elimination delivered the terminal alkene. The findings can be used to predict the regioselectivity in vinylations of other nucleophile classes.

Automated fitting of transition state force fields for biomolecular simulations.

The generation of surrogate potential energy functions (PEF) that are orders of magnitude faster to compute but as accurate as the underlying training data from high-level electronic structure methods is one of the most promising applications of fitting procedures in chemistry. In previous work, we have shown that transition state force fields (TSFFs), fitted to the functional form of MM3* force fields using the quantum guided molecular mechanics (Q2MM) method, provide an accurate description of transition states that can be used for stereoselectivity predictions of small molecule reactions. Here, we demonstrate the applicability of the method for fit TSFFs to the well-established Amber force field, which could be used for molecular dynamics studies of enzyme reaction. As a case study, the fitting of a TSFF to the second hydride transfer in Pseudomonas mevalonii 3-hydroxy-3-methylglutaryl coenzyme A reductase (PmHMGR) is used. The differences and similarities to fitting of small molecule TSFFs are discussed.

Proofreading experimentally assigned stereochemistry through Q2MM predictions in Pd-catalyzed allylic aminations.

The palladium-catalyzed enantioselective allylic substitution by carbon or nitrogen nucleophiles is a key transformation that is particularly useful for the synthesis of bioactive compounds. Unfortunately, the selection of a suitable ligand/substrate combination often requires significant screening effort. Here, we show that a transition state force field (TSFF) derived by the quantum-guided molecular mechanics (Q2MM) method can be used to rapidly screen ligand/substrate combinations. Testing of this method on 77 literature reactions revealed several cases where the computationally predicted major enantiomer differed from the one reported. Interestingly, experimental follow-up led to a reassignment of the experimentally observed configuration. This result demonstrates the power of mechanistically based methods to predict and, where necessary, correct the stereochemical outcome.

Microsecond timescale MD simulations at the transition state of PmHMGR predict remote allosteric residues.

Understanding the mechanisms of enzymatic catalysis requires a detailed understanding of the complex interplay of structure and dynamics of large systems that is a challenge for both experimental and computational approaches. More importantly, the computational demands of QM/MM simulations mean that the dynamics of the reaction can only be considered on a timescale of nanoseconds even though the conformational changes needed to reach the catalytically active state happen on a much slower timescale. Here we demonstrate an alternative approach that uses transition state force fields (TSFFs) derived by the quantum-guided molecular mechanics (Q2MM) method that provides a consistent treatment of the entire system at the classical molecular mechanics level and allows simulations at the microsecond timescale. Application of this approach to the second hydride transfer transition state of HMG-CoA reductase from Pseudomonas mevalonii (PmHMGR) identified three remote residues, R396, E399 and L407, (15-27 Å away from the active site) that have a remote dynamic effect on enzyme activity. The predictions were subsequently validated experimentally via site-directed mutagenesis. These results show that microsecond timescale MD simulations of transition states are possible and can predict rather than just rationalize remote allosteric residues.

Experimental and Computational Models for Side Chain Discrimination in Peptide-Protein Interactions.

A bis(18-crown-6) Tröger's base receptor and 4-substituted hepta-1,7-diyl bisammonium salt ligands have been used as a model system to study the interactions between non-polar side chains of peptides and an aromatic cavity of a protein. NMR titrations and NOESY/ROESY NMR spectroscopy were used to analyze the discrimination of the ligands by the receptor based on the substituent of the ligand, both quantitatively (free binding energies) and qualitatively (conformations). The analysis showed that an all-anti conformation of the heptane chain was preferred for most of the ligands, both free and when bound to the receptor, and that for all of the receptor-ligand complexes, the substituent was located inside or partly inside of the aromatic cavity of the receptor. We estimated the free binding energy of a methyl- and a phenyl group to an aromatic cavity, via CH-π, and combined aromatic CH-π and π-π interactions to be -1.7 and -3.3 kJ mol-1 , respectively. The experimental results were used to assess the accuracy of different computational methods, including molecular mechanics (MM) and density functional theory (DFT) methods, showing that MM was superior.

Stereoselectivity Predictions for the Pd-Catalyzed 1,4-Conjugate Addition Using Quantum-Guided Molecular Mechanics.

The conjugate addition of aryl boronic acids to enones is a powerful synthetic tool to introduce quaternary chiral centers, but the experimentally observed stereoselectivities vary widely, and the identification of suitable substrate-ligand combinations requires significant effort. We describe the development and application of a transition-state force field (TSFF) by the quantum-guided molecular mechanics (Q2MM) method that is validated using an automated screen of 9 ligands, 38 aryl boronic acids, and 22 enones, leading to a MUE of 1.8 kJ/mol and a R2 value of 0.877 over 82 examples. A detailed error analysis identified the structural origin for the deviations in the small group of outliers. The TSFF was then used to predict the stereoselectivity for 27 ligands and 59 enones. The vast majority of the virtual screening results are in line with the expected results. Selected results for 6-substituted pyrox ligands, which were not part of the training set, were followed up by density functional theory and experimental studies.

Machine learning meets mechanistic modelling for accurate prediction of experimental activation energies.

Accurate prediction of chemical reactions in solution is challenging for current state-of-the-art approaches based on transition state modelling with density functional theory. Models based on machine learning have emerged as a promising alternative to address these problems, but these models currently lack the precision to give crucial information on the magnitude of barrier heights, influence of solvents and catalysts and extent of regio- and chemoselectivity. Here, we construct hybrid models which combine the traditional transition state modelling and machine learning to accurately predict reaction barriers. We train a Gaussian Process Regression model to reproduce high-quality experimental kinetic data for the nucleophilic aromatic substitution reaction and use it to predict barriers with a mean absolute error of 0.77 kcal mol-1 for an external test set. The model was further validated on regio- and chemoselectivity prediction on patent reaction data and achieved a competitive top-1 accuracy of 86%, despite not being trained explicitly for this task. Importantly, the model gives error bars for its predictions that can be used for risk assessment by the end user. Hybrid models emerge as the preferred alternative for accurate reaction prediction in the very common low-data situation where only 100-150 rate constants are available for a reaction class. With recent advances in deep learning for quickly predicting barriers and transition state geometries from density functional theory, we envision that hybrid models will soon become a standard alternative to complement current machine learning approaches based on ground-state physical organic descriptors or structural information such as molecular graphs or fingerprints.

Organic reactivity from mechanism to machine learning.

As more data are introduced in the building of models of chemical reactivity, the mechanistic component can be reduced until 'big data' applications are reached. These methods no longer depend on underlying mechanistic hypotheses, potentially learning them implicitly through extensive data training. Reactivity models often focus on reaction barriers, but can also be trained to directly predict lab-relevant properties, such as yields or conditions. Calculations with a quantum-mechanical component are still preferred for quantitative predictions of reactivity. Although big data applications tend to be more qualitative, they have the advantage to be broadly applied to different kinds of reactions. There is a continuum of methods in between these extremes, such as methods that use quantum-derived data or descriptors in machine learning models. Here, we present an overview of the recent machine learning applications in the field of chemical reactivity from a mechanistic perspective. Starting with a summary of how reactivity questions are addressed by quantum-mechanical methods, we discuss methods that augment or replace quantum-based modelling with faster alternatives relying on machine learning.