ElemeNet: Multiscale Molecular Machine Learning with Uncertainty Quantification Across the Periodic Table

ALF: Open-Source Active Learning Framework for Atomistic Modeling

Machine learning interatomic potentials (MLIPs) have surged in popularity over the last two decades, with many model architectures now openly available. As data-driven models, MLIPs critically depend on high-fidelity (i.e., physically accurate) training data produced by electronic structure calculations. However, assembling large and chemically diverse datasets can be a complex and time-consuming endeavor, often requiring the manual selection of representative atomic configurations and the execution of hundreds to millions of electronic structure simulations. To address this challenge, we introduce the Active Learning Framework (ALF), an open-source Python package designed to streamline the design and deployment of MLIP training datasets on High Performance Computing resources. ALF automatically selects new configurations from undersampled regions of the potential energy surface where the MLIP exhibits high uncertainty, schedules electronic structure calculations across available computational resources, and retrains MLIPs on the fly, thereby reducing manual intervention and limiting human bias. As a demonstration, we applied ALF to generate an actively learned dataset for molten salt mixtures consisting of F, Li, Na, Be, and K atoms. An MLIP trained on this data was then employed to predict melting point, viscosity, density, radial distribution function, and specific heat, which are computationally resource-intensive to evaluate via first-principles molecular dynamics. These results were subsequently validated against experimental data. Collectively, these findings illustrate ALF’s effectiveness in compiling datasets that capture essential chemical and structural regimes, thereby virtually eliminating manual curation.

Deep Learning Foundation Models for Low-Data Regimes from Classical Molecular Descriptors

Fast and accurate data-driven prediction of molecular properties is pivotal to scientific advancements across myriad chemical domains. Deep learning methods have recently garnered much attention, despite their inability to outperform classical machine learning methods when tested on practical, real-world benchmarks with limited training data. This study seeks to bridge this gap by introducing a new avenue for foundation model pretraining. We propose pretraining on low-noise, calculable molecular descriptors via supervised learning to obtain rich, highly transferable molecular representations. We demonstrate this strategy with CheMeleon, a O(10M) parameter foundation model that enables directed message-passing neural networks to finally exceed the performance of classical methods in the low-data regime. We evaluate on 58 benchmark data sets spanning a range of properties relevant to small-molecule drug discovery, sourced from the industry-led Polaris benchmarking initiative. Rigorous statistical comparisons show that CheMeleon outperforms classical baselines like Random Forest on molecular fingerprints and descriptors, as well as existing foundation models. We open-source the CheMeleon model and the pretraining framework to encourage adoption and extension of this pretraining strategy across chemical sciences.

How to Build Machine-Learning Models for Molecular Science: A Step-by-Step, Annotated Tutorial

Machine learning (ML) holds immense promise for predictive modeling but is often misapplied without a systematic understanding of the full model development pipeline. This tutorial provides a comprehensive, end-to-end workflow─from raw data to deployed models─explicitly designed for environmental chemists with limited prior experience in ML modeling while also providing practical guidance for other users seeking to strengthen their modeling workflows. Using a public high-performance liquid chromatography small molecule retention time (SMRT) data set containing over 80,000 small molecules, this tutorial demonstrates critical stages: data cleaning via chemical similarity and scaffold analysis, application-oriented data splitting, feature engineering with molecular fingerprints and graphs, model development using commonly used LightGBM and Graph Convolutional Network (GCN), and model interpretation with different methods. This work highlights how domain knowledge guides each step to avoid common pitfalls, such as overestimating model performance through inappropriate data splits. The work concludes with a robust deployment framework that ensures models are both accurate and accessible. All code is openly available, providing an example implementation for users to reference when developing ML models across diverse applications.

MolDeTr: A Chemistry-Informed Deep Learning Model for Next-Generation Automated Analysis of 1H NMR Spectra.

Accurate interpretation of one-dimensional proton nuclear magnetic resonance (1H NMR) spectra remains a rate-limiting step in molecular structure elucidation, particularly when signal overlap, strong spin coupling, and instrumental distortions mask key features. Existing automated approaches depend on computationally intensive and sensitive iterative quantum-mechanical fitting and still require expert oversight. Here we introduce MolDeTr, a chemistry-informed deep-learning framework derived from the detection-transformer architecture that unifies peak picking, multiplet identification, and extraction of chemical shifts, scalar coupling constants, relaxation-dependent decay times, and proton counts in a single-network pass. The method targets prototypical spin systems of single-component small molecules in 1D 1H NMR, with up to ten distinct groups of chemically equivalent spins (multiplets). MolDeTr is trained exclusively on synthetic spectra generated by spin-dynamics simulations and augmented with realistic experimental artifacts, enabling it to generalize to unseen compounds─including experimental spectra with overlapping and strongly coupled multiplets─without reference standards or prior spin-system knowledge. Unlike structure-conditioned shift-prediction or calculation models, e.g., density functional theory (DFT), that assume the molecular structure is known, MolDeTr addresses the spectrum-conditioned inverse problem and extracts spin-system parameters directly from measured 1D 1H NMR spectra, thereby substantially improving chemical-shift prediction precision by one to 2 orders of magnitude compared to existing structure-conditioned approaches. Benchmarking against a diverse experimental set of 1H NMR spectra of modestly sized small molecules, spanning 80 to 600 MHz base frequency, shows median absolute errors of 0.89 Hz for chemical shifts and 0.20 Hz for coupling constants, while absolute proton counts are predicted with 93.5% accuracy, outperforming state-of-the-art spectrum analysis software and experienced spectroscopists. By eliminating iterative fitting and expert intervention, MolDeTr offers a scalable route to fully automated spectral analysis, accelerating molecular discovery across the chemical sciences.

Robust generative transition-state models for unseen chemistry.

Fast and Accurate Foundation Models for Equivariant Machine-Learned Interatomic Potentials

Machine-learned interatomic potentials (MLIPs) have emerged as a transformative tool for computational materials science and chemistry, with universal potentials trained on large and diverse datasets now routinely deployed as'foundation models'for downstream fine-tuning in targeted chemical spaces. Many scientific applications of the resulting models, such as molecular dynamics (MD), require high inference and training speeds as well as accuracy. In this work, we examine the limits of equivariant MLIPs, which directly encode physical symmetries in model architectures, to achieve these competing targets -- particularly in the regime of extremely large datasets where data efficiency is less critical. We show how this trade-off can be addressed, and present a family of foundation potentials in the NequIP and Allegro equivariant MLIP architectures which achieve leading inference speeds and strong scalability as well as excellent accuracies across a range of community benchmarks -- spanning materials discovery, thermal conductivity prediction, and near-equilibrium mechanical and thermodynamic properties. Accelerations implemented within the NequIP infrastructure now permit training of high-accuracy foundation potentials on ultra-large datasets with dramatically reduced computational cost. Alongside, we show that efforts to improve model accuracy for materials discovery should focus on dataset diversity and improved, consistent descriptions of transition metal compound energy surfaces.