Mechanistic modeling of chemical transformations offers a compelling basis for understanding reactivity and allows for prediction of reaction outcomes before attempting experiments. Despite progress in machine learned interatomic potentials (MLIPs), we demonstrate that available models lack the accuracy for diverse reaction modeling. With this motivation, we developed a general MLIP for mechanistic modeling of organics, AIMNet2-rxn, using a dataset of ~4.7 x 106 range-separated DFT calculations. AIMNet2-rxn enables reaction modeling ~106 faster than the reference quantum mechanical (QM) methods while significantly outperforming graph-based ML, reaffirming the value using 3D chemical information for training. On a test suite of well-known reaction mechanisms—such as amide formation, proton transfers, and pericyclics—AIMNet2-rxn yields 1-2 kcal mol-1 accuracy across reaction coordinates without retraining or system-specific fine-tuning. To exploit GPU parallelism and AIMNet2-rxn efficiency, we introduce a batched nudged elastic band (BNEB) method that readily achieves minimum energy pathway search on a millions-of-reactions scale. To demonstrate complex reaction characterization, the thermodynamics of an 11-step pathway producing hydroxymethylfurfural, the experimentally observed major product of glucose pyrolysis, is evaluated. Overall, the accuracy and efficiency afforded by AIMNet2-rxn creates opportunities in high-throughput reaction discovery and deep reaction network analysis that would be infeasible with QM methods.

Finding efficient substrate-catalyst combinations for palladium-catalyzed cross-coupling reactions remains a critical challenge in synthetic chemistry, with broad implications for pharmaceutical and materials manufacturing. We report AIMNet2-Pd, a machine learned interatomic potential that enables rapid, accurate computational studies of palladium-catalyzed cross-coupling reactions. AIMNet2-Pd replaces computationally expensive electronic structure calculations with a neural network-based model that performs geometry optimization, transition state searches, and energy calculations in seconds while maintaining accuracy within 1-2 kcal mol⁻¹ and ~0.1 Å compared to the reference QM calculations. AIMNet2-Pd makes computational high-throughput catalyst screening and mechanistic studies of realistic systems feasible by providing on-demand thermodynamic and kinetic predictions for each step of a catalytic cycle. Importantly, the applicability of the systems extends beyond the monophosphine ligands in Pd(0)/Pd(II) cycles for which it has been trained on to chemically diverse Pd complexes. This demonstrates AIMNet2-Pd's utility to serve as a general-purpose and high-throughput tool for studying catalytic reactions.

Journal: Chemical Science

Machine learned interatomic potentials (MLIPs) are reshaping computational chemistry practices because of their ability to drastically exceed the accuracy-length/time scale tradeoff.

Journal: Journal of Chemical Information and Modeling

Proteochemometric models (PCM) are used in computational drug discovery to leverage both protein and ligand representations for bioactivity prediction. While machine learning (ML) and deep learning (DL) have come to dominate PCMs, often serving as scoring functions, rigorous evaluation standards have not always been consistently applied. In this study, using kinase-ligand bioactivity prediction as a model system, we highlight the critical roles of dataset curation, permutation testing, class imbalances, data splitting strategies, and embedding quality in determining model performance. Our findings indicate that data splitting and class imbalances are the most critical factors affecting PCM performance, emphasizing the challenges in generalizing ability of ML/DL-PCMs. We evaluated various protein-ligand descriptors and embeddings, including those augmented with multiple sequence alignment (MSA) information. However, permutation testing consistently demonstrated that protein embeddings contributed minimally to PCM efficacy. This study advocates for the adoption of stringent evaluation standards to enhance the generalizability of models to out-of-distribution data and improve benchmarking practices.

We present an innovative methodology for measuring amide coupling reaction rates by monitoring pH changes via indicator dyes, achieving precision comparable to traditional NMR techniques, called PRISM (Parallelized Reaction-rates via Indicator Spectrometry using Machine-vision) The experimental design, enabled by a serial dilution, allowed for measuring twelve rate constants concurrently, spanning more than four orders of magnitude using 96-well plates, with 1,162 total rate constants collected. Moreover, the instrumentation is 3D-printed, with the remaining components comprising readily available and cost-effective hardware, promoting the democratized use of this technique to generate uniform data sets. Validation with 19F-NMR confirmed PRISM’s reliability. Computational investigations reveal a concerted asynchronous SN2 mechanism, with base-catalyzed pathways exhibiting the lowest energy barriers. To complement the PRISM rate dataset, we developed a classification model that achieves high accuracy for out-of-distribution reactants in determining rate measurability, and a chemically rich graph neural network regression model for predicting quantitative reaction rates. This approach provides a framework that offers a resource-efficient strategy for studying reaction kinetics, which can be applied to other reaction classes.

Journal: Journal of Chemical Theory and Computation

The AIQM series methods are successful neural network-based models that target coupled-cluster accuracy while maintaining high robustness and transferability across various tasks by leveraging Δ-learning. However, the previous AIQM1 and AIQM2 models are limited to molecular systems with four elements: H, C, N, and O, which falls short of meeting the common needs for atomistic simulations. Here, we introduce the extension—AIQM3—that covers three additional chemical elements: S, F, Cl, and approaches coupled cluster level at the speed of a semi-empirical method. AIQM3 maintains the accuracy of its predecessor AIQM2, surpasses the commonly used density functional theory (DFT) method in different types of molecular interactions, and its efficiency is competitive with that of machine learning interatomic potentials on commodity CPU hardware. AIQM3 superiority is showcased for reaction simulations and tasks related to drug design, where it delivers accurate torsion profiles for various real-world drug-like molecules. In addition, AIQM3 can be used for infrared (IR) spectra calculations at a low cost. We provide a web service for the AIQM3 calculations on the Aitomistic Hub at aitomistic.xyz, to democratize and facilitate its use with the assistance of AI agents.

Journal: Proceedings of the 34th ACM International Conference on Information and Knowledge Management

Automation of experiments in cloud laboratories promises to revolutionize scientific research by enabling remote experimentation and improving reproducibility. However, maintaining quality control without constant human oversight remains a critical challenge. Here, we present a novel machine learning framework for automated anomaly detection in High-Performance Liquid Chromatography (HPLC) experiments conducted in a cloud lab. Our system specifically targets air bubble contamination—a common yet challenging issue that typically requires expert analytical chemists to detect and resolve. By leveraging active learning combined with human-in-the-loop annotation, we trained a binary classifier on approximately 25,000 HPLC traces. Prospective validation demonstrated robust performance, with an accuracy of 0.96 and an F1 score of 0.92, suitable for real-world applications. Beyond anomaly detection, we show that the system can serve as a sensitive indicator of instrument health, outperforming traditional periodic qualification tests in identifying systematic issues. The framework is protocol-agnostic, instrument-agnostic, and vendor-neutral, making it adaptable to various laboratory settings. This work represents a significant step toward fully autonomous laboratories by enabling continuous quality control, reducing the expertise barrier for complex analytical techniques, and facilitating proactive maintenance of scientific instrumentation. The approach can be extended to detect other types of experimental anomalies, potentially transforming how quality control is implemented in self-driving laboratories (SDLs) across diverse scientific disciplines.

Journal: Digital Discovery

Autonomous experiments are vulnerable to unforeseen adverse events. We developed a transferable ML framework that flags affected HPLC runs in real time and provides expert-level quality control without human oversight.

Journal: Nature Computational Science

Journal: Angewandte Chemie International Edition

Abstract Open‐shell systems such as radical intermediates are central to radical polymerization (RP), combustion, catalysis, and many other chemical and industrial processes, yet their accurate modeling presents significant computational challenges. Most of the current machine learning interatomic potentials do not distinguish between different spin states, making them unsuitable for open‐shell reactive chemistry. Here we present AIMNet2‐NSE (neural spin‐charge equilibration), a neural network potential that incorporates spin‐charge equilibration for accurate treatment of molecules and reactions with arbitrary charge and spin multiplicities. Built upon the AIMNet2 framework, AIMNet2‐NSE is trained on an extensive dataset comprising 20 million closed‐shell neutral and charged molecules, 13 million open‐shell radical configurations, and 200K radical reaction profiles. With explicit handling of spin charges, AIMNet2‐NSE enables prediction of spin‐resolved properties with near‐DFT accuracy while maintaining a favorable linear scaling compared to the polynomial scaling of electronic structure methods. The predictive capabilities and generalizability of our model are confirmed by evaluations on large‐scale radical test sets, the industrially relevant BASChem19 benchmark, and RP reactions. Overall, AIMNet2‐NSE represents a significant advancement in machine learning interatomic potentials, allowing efficient exploration of complex open‐shell systems, and significantly advancing our ability to model radical reaction pathways and reactive intermediates in chemical processes where traditional quantum mechanical methods are computationally prohibitive.

Journal: Angewandte Chemie

Abstract Open‐shell systems such as radical intermediates are central to radical polymerization (RP), combustion, catalysis, and many other chemical and industrial processes, yet their accurate modeling presents significant computational challenges. Most of the current machine learning interatomic potentials do not distinguish between different spin states, making them unsuitable for open‐shell reactive chemistry. Here we present AIMNet2‐NSE (neural spin‐charge equilibration), a neural network potential that incorporates spin‐charge equilibration for accurate treatment of molecules and reactions with arbitrary charge and spin multiplicities. Built upon the AIMNet2 framework, AIMNet2‐NSE is trained on an extensive dataset comprising 20 million closed‐shell neutral and charged molecules, 13 million open‐shell radical configurations, and 200K radical reaction profiles. With explicit handling of spin charges, AIMNet2‐NSE enables prediction of spin‐resolved properties with near‐DFT accuracy while maintaining a favorable linear scaling compared to the polynomial scaling of electronic structure methods. The predictive capabilities and generalizability of our model are confirmed by evaluations on large‐scale radical test sets, the industrially relevant BASChem19 benchmark, and RP reactions. Overall, AIMNet2‐NSE represents a significant advancement in machine learning interatomic potentials, allowing efficient exploration of complex open‐shell systems, and significantly advancing our ability to model radical reaction pathways and reactive intermediates in chemical processes where traditional quantum mechanical methods are computationally prohibitive.

Journal: JACS Au

Journal: Crystal Growth & Design

Molecular geometry is crucial for biological activity and chemical reactivity; however, computational methods for generating 3D structures are limited by the vast scale of conformational space and the complexities of stereochemistry. Here we present an approach that combines an expansive dataset of molecular conformers with generative diffusion models to address this problem. We introduce ChEMBL3D, which contains over 250 million molecular geometries for 1.8 million drug-like compounds, optimized using AIMNet2 neural network potentials to a near-quantum mechanical accuracy with implicit solvent effects included. This dataset captures complex organic molecules in various protonation states and stereochemical configurations. We then developed LoQI, a stereochemistry-aware diffusion model that learns molecular geometry distributions directly from this data. Through graph augmentation, LoQI accurately generates molecular structures with targeted stereochemistry, representing a significant advance in modeling capabilities over previous generative methods. The model outperforms traditional approaches, achieving up to tenfold improvements in energy accuracy and effective recovery of optimal conformations. Benchmark tests on complex systems, including macrocycles and flexible molecules, as well as validation with crystal structures, show LoQI can perform low energy conformer search efficiently. The model code and dataset are available at https: //github.com/isayevlab/LoQI.

Journal: Digital Discovery

Revisiting GEOM drugs: corrected metrics and novel energy-based structural benchmark enable rigorous evaluation of 3D molecule generative models.

Journal: Angewandte Chemie

Abstract The development of high‐performance elastomers for additive manufacturing requires overcoming complex property trade‐offs that challenge conventional material discovery pipelines. Here, a human‐in‐the‐loop reinforcement learning (RL) approach is used to discover polyurethane elastomers that overcome pervasive stress–strain property tradeoffs. Starting with a diverse training set of 92 formulations, a coupled multi‐component reward system was identified that guides RL agents toward materials with both high strength and extensibility. Through three rounds of iterative optimization combining RL predictions with human chemical intuition, we identified elastomers with more than double the average toughness compared to the initial training set. The final exploitation round, aided by solubility prescreening, predicted twelve materials exhibiting both high strength (>10 MPa) and high strain at break (>200%). Analysis of the high‐performing materials revealed structure‐property insights, including the benefits of high molar mass urethane oligomers, a high density of urethane functional groups, and incorporation of rigid low molecular weight diols and unsymmetric diisocyanates. These findings demonstrate that machine‐guided, human‐augmented design is a powerful strategy for accelerating polymer discovery in applications where data is scarce and expensive to acquire, with broad applicability to multi‐objective materials optimization.

Journal: Angewandte Chemie International Edition

Abstract The development of high‐performance elastomers for additive manufacturing requires overcoming complex property trade‐offs that challenge conventional material discovery pipelines. Here, a human‐in‐the‐loop reinforcement learning (RL) approach is used to discover polyurethane elastomers that overcome pervasive stress–strain property tradeoffs. Starting with a diverse training set of 92 formulations, a coupled multi‐component reward system was identified that guides RL agents toward materials with both high strength and extensibility. Through three rounds of iterative optimization combining RL predictions with human chemical intuition, we identified elastomers with more than double the average toughness compared to the initial training set. The final exploitation round, aided by solubility prescreening, predicted twelve materials exhibiting both high strength (>10 MPa) and high strain at break (>200%). Analysis of the high‐performing materials revealed structure‐property insights, including the benefits of high molar mass urethane oligomers, a high density of urethane functional groups, and incorporation of rigid low molecular weight diols and unsymmetric diisocyanates. These findings demonstrate that machine‐guided, human‐augmented design is a powerful strategy for accelerating polymer discovery in applications where data is scarce and expensive to acquire, with broad applicability to multi‐objective materials optimization.