Comprehensive exploration of graphically defined reaction spaces

Uncertainty Qualification for Deep Learning-Based Elementary Reaction Property Prediction

The prediction of the thermodynamic and kinetic properties of elementary reactions has shown rapid improvement due to the implementation of deep learning (DL) methods. While various studies have reported the success in predicting reaction properties, the quantification of prediction uncertainty has seldom been investigated, thus compromising the confidence in using these predicted properties in practical applications. Here, we integrated graph convolutional neural networks (GCNN) with three uncertainty prediction techniques, including deep ensemble, Monte Carlo (MC)-dropout, and evidential learning, to provide insights into the uncertainty quantification and utility. The deep ensemble model outperforms others in accuracy and shows the highest reliability in estimating prediction uncertainty across all elementary reaction property data sets. We also verified that the deep ensemble model showed a satisfactory capability in recognizing epistemic and aleatoric uncertainties. Additionally, we adopted a Monte Carlo Tree Search method for extracting the explainable reaction substructures, providing a chemical explanation for DL predicted properties and corresponding uncertainties. Finally, to demonstrate the utility of uncertainty qualification in practical applications, we performed an uncertainty-guided calibration of the DL-constructed kinetic model, which achieved a 25% higher hit ratio in identifying dominant reaction pathways compared to that of the calibration without uncertainty guidance.

Reproducing Reaction Mechanisms with Machine-Learning Models Trained on a Large-Scale Mechanistic Dataset

Mechanistic understanding of organic reactions can facilitate reaction development, impurity prediction, and in principle, reaction discovery. While several machine learning models have sought to address the task of predicting reaction products, their extension to predicting reaction mechanisms has been impeded by the lack of a corresponding mechanistic dataset. In this study, we construct such a dataset by imputing intermediates between experimentally reported reactants and products using expert reaction templates and train several machine learning models on the resulting dataset of 5,184,184 elementary steps. We explore the performance and capabilities of these models, focusing on their ability to predict reaction pathways and recapitulate the roles of catalysts and reagents. Additionally, we demonstrate the potential of mechanistic models in predicting impurities, often overlooked by conventional models. We conclude by evaluating the generalizability of mechanistic models to new reaction types, revealing challenges related to dataset diversity, consecutive predictions, and violations of atom conservation.

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.

Accurately Predicting Barrier Heights for Radical Reactions in Solution Using Deep Graph Networks

Quantitative estimates of reaction barriers and solvent effects are essential for developing kinetic mechanisms and predicting reaction outcomes. Here, we create a new data set of 5,600 unique elementary radical reactions calculated using the M06-2X/def2-QZVP//B3LYP-D3(BJ)/def2-TZVP level of theory. A conformer search is done for each species using TPSS/def2-TZVP. Gibbs free energies of activation and of reaction for these radical reactions in 40 common solvents are obtained using COSMO-RS for solvation effects. These balanced reactions involve the elements H, C, N, O, and S, contain up to 19 heavy atoms, and have atom-mapped SMILES. All transition states are verified by an intrinsic reaction coordinate calculation. We next train a deep graph network to directly estimate the Gibbs free energy of activation and of reaction in both gas and solution phases using only the atom-mapped SMILES of the reactant and product and the SMILES of the solvent. This simple input representation avoids computationally expensive optimizations for the reactant, transition state, and product structures during inference, making our model well-suited for high-throughput predictive chemistry and quickly providing information for (retro-)synthesis planning tools. To properly measure model performance, we report results on both interpolative and extrapolative data splits and also compare to several baseline models. During training and testing, the data set is augmented by including the reverse direction of each reaction and variants with different resonance structures. After data augmentation, we have around 2 million entries to train the model, which achieves a testing set mean absolute error of 1.16 kcal mol-1 for the Gibbs free energy of activation in solution. We anticipate this model will accelerate predictions for high-throughput screening to quickly identify relevant reactions in solution, and our data set will serve as a benchmark for future studies.

Reactant Discovery with an Ab Initio Nanoreactor: Exploration of Astrophysical N-Heterocycle Precursors and Formation Pathways

The incorporation of nitrogen atoms into cyclic compounds is essential for terrestrial life; nitrogen-containing (N-)heterocycles make up DNA and RNA nucleobases, several amino acids, B vitamins, porphyrins, and other components of biomolecules. The discovery of these molecules on meteorites with non-terrestrial isotopic abundances supports the hypothesis of exogenous delivery of prebiotic material to early Earth; however, there has been no detection of these species in interstellar environments, indicating that there is a need for greater knowledge of their astrochemical formation and destruction pathways. Here, we present results of simulations of gas-phase pyrrole and pyridine formation from an ab initio nanoreactor, a first-principles molecular dynamics simulation method that accelerates reaction discovery by applying non-equilibrium forces that are agnostic to individual reaction coordinates. Using the nanoreactor in a retrosynthetic mode, starting with the N-heterocycle of interest and a radical leaving group, then considering the discovered reaction pathways in reverse, a rich landscape of N-heterocycle-forming reactivity can be found. Several of these reaction pathways, when mapped to their corresponding minimum energy paths, correspond to novel barrierless formation pathways for pyridine and pyrrole, starting from both detected and hypothesized astrochemical precursors. This study demonstrates how first-principles reaction discovery can build mechanistic knowledge in astrochemical environments as well as in early Earth models such as Titan's atmosphere where N-heterocycles have been tentatively detected.

Machine Learning of Reactive Potentials

In the past two decades, machine learning potentials (MLPs) have driven significant developments in chemical, biological, and material sciences. The construction and training of MLPs enable fast and accurate simulations and analysis of thermodynamic and kinetic properties. This review focuses on the application of MLPs to reaction systems with consideration of bond breaking and formation. We review the development of MLP models, primarily with neural network and kernel-based algorithms, and recent applications of reactive MLPs (RMLPs) to systems at different scales. We show how RMLPs are constructed, how they speed up the calculation of reactive dynamics, and how they facilitate the study of reaction trajectories, reaction rates, free energy calculations, and many other calculations. Different data sampling strategies applied in building RMLPs are also discussed with a focus on how to collect structures for rare events and how to further improve their performance with active learning.

Benchmarking machine-readable vectors of chemical reactions on computed activation barriers

In recent years, there has been a surge of interest in predicting computed activation barriers, to enable the acceleration of the automated exploration of reaction networks. Consequently, various predictive approaches have emerged, ranging from graph-based models to methods based on the three-dimensional structure of reactants and products. In tandem, many representations have been developed to predict experimental targets, which may hold promise for barrier prediction as well. Here, we bring together all of these efforts and benchmark various methods (Morgan fingerprints, the DRFP, the CGR representation-based Chemprop, SLATMd, B2Rl2, EquiReact and language model BERT + RXNFP) for the prediction of computed activation barriers on three diverse datasets.

Organic Reactivity Made Easy and Accurate with Automated Multireference Calculations

In organic reactivity studies, quantum chemical calculations play a pivotal role as the foundation of understanding and machine learning model development. While prevalent black-box methods like density functional theory (DFT) and coupled-cluster theory (e.g., CCSD(T)) have significantly advanced our understanding of chemical reactivity, they frequently fall short in describing multiconfigurational transition states and intermediates. Achieving a more accurate description necessitates the use of multireference methods. However, these methods have not been used at scale due to their often-faulty predictions without expert input. Here, we overcome this deficiency with automated multiconfigurational pair-density functional theory (MC-PDFT) calculations. We apply this method to 908 automatically generated organic reactions. We find 68% of these reactions present significant multiconfigurational character in which the automated multiconfigurational approach often provides a more accurate and/or efficient description than DFT and CCSD(T). This work presents the first high-throughput application of automated multiconfigurational methods to reactivity, enabled by automated active space selection algorithms and the computation of electronic correlation with MC-PDFT on-top functionals. This approach can be used in a black-box fashion, avoiding significant active space inconsistency error in both single- and multireference cases and providing accurate multiconfigurational descriptions when needed.

Graph to Activation Energy Models Easily Reach Irreducible Errors but Show Limited Transferability

Activation energy characterization of competing reactions is a costly but crucial step for understanding the kinetic relevance of distinct reaction pathways, product yields, and myriad other properties of reacting systems. The standard methodology for activation energy characterization has historically been a transition state search using the highest level of theory that can be afforded. However, recently, several groups have popularized the idea of predicting activation energies directly based on nothing more than the reactant and product graphs, a sufficiently complex neural network, and a broad enough data set. Here, we have revisited this task using the recently developed Reaction Graph Depth 1 (RGD1) transition state data set and several newly developed graph attention architectures. All of these new architectures achieve similar state-of-the-art results of ∼4 kcal/mol mean absolute error on withheld testing sets of reactions but poor performance on external testing sets composed of reactions with differing mechanisms, reaction molecularity, or reactant size distribution. Limited transferability is also shown to be shared by other contemporary graph to activation energy architectures through a series of case studies. We conclude that an array of standard graph architectures can already achieve results comparable to the irreducible error of available reaction data sets but that out-of-distribution performance remains poor.

Diffusion-based generative AI for exploring transition states from 2D molecular graphs

The exploration of transition state (TS) geometries is crucial for elucidating chemical reaction mechanisms and modeling their kinetics. Recently, machine learning (ML) models have shown remarkable performance for prediction of TS geometries. However, they require 3D conformations of reactants and products often with their appropriate orientations as input, which demands substantial efforts and computational cost. Here, we propose a generative approach based on the stochastic diffusion method, namely TSDiff, for prediction of TS geometries just from 2D molecular graphs. TSDiff outperforms the existing ML models with 3D geometries in terms of both accuracy and efficiency. Moreover, it enables to sample various TS conformations, because it learns the distribution of TS geometries for diverse reactions in training. Thus, TSDiff finds more favorable reaction pathways with lower barrier heights than those in the reference database. These results demonstrate that TSDiff shows promising potential for an efficient and reliable TS exploration.

Accurate transition state generation with an object-aware equivariant elementary reaction diffusion model

Transition state search is key in chemistry for elucidating reaction mechanisms and exploring reaction networks. The search for accurate 3D transition state structures, however, requires numerous computationally intensive quantum chemistry calculations due to the complexity of potential energy surfaces. Here we developed an object-aware SE(3) equivariant diffusion model that satisfies all physical symmetries and constraints for generating sets of structures-reactant, transition state and product-in an elementary reaction. Provided reactant and product, this model generates a transition state structure in seconds instead of hours, which is typically required when performing quantum-chemistry-based optimizations. The generated transition state structures achieve a median of 0.08 Å root mean square deviation compared to the true transition state. With a confidence scoring model for uncertainty quantification, we approach an accuracy required for reaction barrier estimation (2.6 kcal mol-1) by only performing quantum chemistry-based optimizations on 14% of the most challenging reactions. We envision usefulness for our approach in constructing large reaction networks with unknown mechanisms.

Δ2 machine learning for reaction property prediction

The emergence of Δ-learning models, whereby machine learning (ML) is used to predict a correction to a low-level energy calculation, provides a versatile route to accelerate high-level energy evaluations at a given geometry. However, Δ-learning models are inapplicable to reaction properties like heats of reaction and activation energies that require both a high-level geometry and energy evaluation. Here, a Δ2-learning model is introduced that can predict high-level activation energies based on low-level critical-point geometries. The Δ2 model uses an atom-wise featurization typical of contemporary ML interatomic potentials (MLIPs) and is trained on a dataset of ∼167 000 reactions, using the GFN2-xTB energy and critical-point geometry as a low-level input and the B3LYP-D3/TZVP energy calculated at the B3LYP-D3/TZVP critical point as a high-level target. The excellent performance of the Δ2 model on unseen reactions demonstrates the surprising ease with which the model implicitly learns the geometric deviations between the low-level and high-level geometries that condition the activation energy prediction. The transferability of the Δ2 model is validated on several external testing sets where it shows near chemical accuracy, illustrating the benefits of combining ML models with readily available physical-based information from semi-empirical quantum chemistry calculations. Fine-tuning of the Δ2 model on a small number of Gaussian-4 calculations produced a 35% accuracy improvement over DFT activation energy predictions while retaining xTB-level cost. The Δ2 model approach proves to be an efficient strategy for accelerating chemical reaction characterization with minimal sacrifice in prediction accuracy.

The challenge of balancing model sensitivity and robustness in predicting yields: a benchmarking study of amide coupling reactions

Accurate prediction of reaction yield is the holy grail for computer-assisted synthesis prediction, but current models have failed to generalize to large literature datasets. To understand the causes and inspire future design, we systematically benchmarked the yield prediction task. We carefully curated and augmented a literature dataset of 41 239 amide coupling reactions, each with information on reactants, products, intermediates, yields, and reaction contexts, and provided 3D structures for the molecules. We calculated molecular features related to 2D and 3D structure information, as well as physical and electronic properties. These descriptors were paired with 4 categories of machine learning methods (linear, kernel, ensemble, and neural network), yielding valuable benchmarks about feature and model performance. Despite the excellent performance on a high-throughput experiment (HTE) dataset (R2 around 0.9), no method gave satisfactory results on the literature data. The best performance was an R2 of 0.395 ± 0.020 using the stack technique. Error analysis revealed that reactivity cliff and yield uncertainty are among the main reasons for incorrect predictions. Removing reactivity cliffs and uncertain reactions boosted the R2 to 0.457 ± 0.006. These results highlight that yield prediction models must be sensitive to the reactivity change due to the subtle structure variance, as well as be robust to the uncertainty associated with yield measurements.

Deep reaction network exploration of glucose pyrolysis

Resolving the reaction networks associated with biomass pyrolysis is central to understanding product selectivity and aiding catalyst design to produce more valuable products. However, even the pyrolysis network of relatively simple [Formula: see text]-D-glucose remains unresolved due to its significant complexity in terms of the depth of the network and the number of major products. Here, a transition-state-guided reaction exploration has been performed that provides complete pathways to most significant experimental pyrolysis products of [Formula: see text]-D-glucose. The resulting reaction network involves over 31,000 reactions and transition states computed at the semiempirical quantum chemistry level and approximately 7,000 kinetically relevant reactions and transition states characterized with density function theory, comprising the largest reaction network reported for biomass pyrolysis. The exploration was conducted using graph-based rules to explore the reactivities of intermediates and an adaption of the Dijkstra algorithm to identify kinetically relevant intermediates. This simple exploration policy surprisingly (re)identified pathways to most major experimental pyrolysis products, many intermediates proposed by previous computational studies, and also identified new low-barrier reaction mechanisms that resolve outstanding discrepancies between reaction pathways and yields in isotope labeling experiments. This network also provides explanatory pathways for the high yield of hydroxymethylfurfural and the reaction pathway that contributes most to the formation of hydroxyacetaldehyde during glucose pyrolysis. Due to the limited domain knowledge required to generate this network, this approach should also be transferable to other complex reaction network prediction problems in biomass pyrolysis.

Accelerating Reaction Network Explorations with Automated Reaction Template Extraction and Application

Autonomously exploring chemical reaction networks with first-principles methods can generate vast data. Especially autonomous explorations without tight constraints risk getting trapped in regions of reaction networks that are not of interest. In many cases, these regions of the networks are only exited once fully searched. Consequently, the required human time for analysis and computer time for data generation can make these investigations unfeasible. Here, we show how simple reaction templates can facilitate the transfer of chemical knowledge from expert input or existing data into new explorations. This process significantly accelerates reaction network explorations and improves cost-effectiveness. We discuss the definition of the reaction templates and their generation based on molecular graphs. The resulting simple filtering mechanism for autonomous reaction network investigations is exemplified with a polymerization reaction.