Aug 2026· Molecules· Vol 31· 0 citations· 86 references
Medicine
TL;DR
This work not only establishes a pioneering paradigm for interpretable ML-driven force field refinement but also provides the first feature engineering solution incorporating chemical, physical, and structural information specifically designed for the machine learning of energetic molecular crystals.
Abstract
Accurate energy prediction in organic molecular crystals, such as cyclotrimethylenetrinitramine (RDX) and cyclotetramethylenetetranitramine (HMX), is hindered by the scarcity of descriptors capable of encoding their hierarchical architecture. Furthermore, while reactive force field (ReaxFF) offers a pathway to simulating chemical reactions, its accuracy for polymorph stability remains suboptimal, and machine learning (ML) driven refinement of ReaxFF for molecular systems remains unexplored. Herein, we report the first construction of a Hierarchical “Center-Environment” (HCE) feature framework specifically tailored for molecular crystals. The HCE framework hierarchically decomposes structural complexity into intramolecular (ring vs. nitro groups) and intermolecular (central molecule vs. coordination shell) attributes, integrating physics-based priors with distance-weighted attention. Using this compact descriptor set, we present the first attempt to rectify ReaxFF predictions via ML, benchmarking against 5930 RDX and 3335 HMX DFT-calculated energies. Our models achieve a significant reduction in mean absolute error: from ~116.7 to 42.2 meV/atom for RDX (kernel ridge regression, KRR) and from 112.3 to 53.7 meV/atom for HMX (support vector regression, SVR). Crucially, HCE features enable robust cross-molecule transferability; a neural network pre-trained on RDX yields a validation error of 52.9 meV/atom on HMX, surpassing models trained exclusively on HMX data. This work not only establishes a pioneering paradigm for interpretable ML-driven force field refinement but also provides the first feature engineering solution incorporating chemical, physical, and structural information specifically designed for the machine learning of energetic molecular crystals.
A machine learning approach is presented that accelerates DFTB simulations by predicting optimal initial atomic charges and demonstrates that ML-predicted initial charges consistently and significantly improve SCC convergence across diverse chemical systems including organic molecules, biomolecules, water clusters, transition metal oxides and solid electrolytes.
Maximilian L. Ach, Karsten Reuter, C. Panosetti· 0 citations
GeoNet is a physicochemical-principle-guided framework for modeling dual-range atomic interactions that achieves the smallest model size and the shortest training time, demonstrating both superior predictive performance and computational efficiency.
A data-driven framework combining explainable machine learning (ML) with large-scale virtual library generation with large-scale virtual library generation is presented, establishing a practical route from experimental data to actionable catalyst designs.
Xuefeng Li, Haoke Qiu, Hanwen Pei et al.· Journal of Physical Chemistr...· 0 citations
Characterizing the solvation of open-shell transition metal ions remains a challenge for classical force fields due to ligand field electronic effects. Here, we present AMOEBA + NN, a machine-learning-augmented polarizable potential, to investigate Cu2+ solvation in NH3 and H2O and mixed environments. The model, trained on quantum-mechanical (QM) association energies, accurately reproduces the Cu2+ energy landscape across diverse coordination geometries. Molecular dynamics simulations demonstrate that AMOEBA + NN successfully captures the electronically driven Jahn–Teller (JT) distortion. Evidence from radial distribution functions (RDFs) and geometry optimizations reveals characteristic axial elongation, which is absent in conventional classical descriptions. Kinetic analysis shows a highly dynamic Cu2+–NH3 environment with a rapid ligand exchange (residence time of 66.52 ps), contrasting with the H2O system where binding is two to three orders of magnitude more stable. Furthermore, the calculated hydration free energy of −477.59 ± 0.50 kcal mol−1 shows excellent agreement with experimental data (within 1.31 kcal mol−1). This work provides a unified, computationally efficient framework for describing the structural, dynamic, and thermodynamic observables of transition metal coordination.
Zhecheng He, Yanxing Wang, Shubham Chatterjee et al.· Chemical Science· 0 citations
The process of drug unbinding is of immense importance in the field of biophysics and therapeutics. The behavior of these systems is greatly influenced by their thermodynamic and kinetic properties. Therefore, it is crucial to accurately estimate the ligand binding free energies and rate of ligand dissociation, yet these processes are often governed by rare event transitions that lie beyond the reach of standard brute-force molecular dynamics simulations. While enhanced sampling simulations offer a solution, their efficacy is strictly contingent upon the selection of appropriate collective variables (CVs) which is non-trivial for complex systems like protein-ligand complexes. In this study, we present a method to derive optimized CV from transition state region (TS) via an interpretable machine learning (ML) model, Elastic Net. By employing some physically intuitive order parameters, the derived optimized CV from the TS-region greatly accelerate ligand binding-unbinding transitions and achieves rapid free energy surface (FES) convergence across diverse systems including buried and solvent exposed active sites such as Trpsin-benzamidine complex, host-guest systems and sodium epoxidase etc. Intriguingly significant contribution of the ligand hydration is found in the optimized CV which depicts crucial role of solvent in driving ligand binding-unbinding transitions. The estimated binding free energies for different protein-ligand complexes match quite well with experiments, while maintaining a low computational cost. The derived optimized CV is also used to calculate the ligand residence times across different systems and calculated residence times are within the experimental range for all systems, again with very little computational costs. Moreover, we show that the optimized CV constructed from TS region via an interpretable ML model is transferable across diverse systems, offering a robust and scalable framework for drug discovery and investigation of complex biomolecular recognition.
This work demonstrates how recent foundational machine learning interatomic potentials (MLIPs) trained at the r$^2$SCAN level can be leveraged to improve the agreement of formation energies with experiment, reducing the mean absolute error by more than 40% relative to GGA without requiring any additional DFT calculation.
Timo Reents, Marnik Bercx, Giovanni Pizzi· 0 citations