CE-ViSNet: An Equivariant Geometric Deep Learning Molecular Force Field with Explicit Chirality Representation
DOI:
https://doi.org/10.70088/vz8rsx64Keywords:
ViSNet, CE-ViSNet, Equivariant graph neural networks, Chiral molecules, Molecular force fieldsAbstract
The molecular equivariant graph neural network ViSNet is a state-of-the-art model in molecular geometric deep learning. However, because it extracts mirror-invariant geometric features solely through vector dot products, it faces challenges in distinguishing enantiomers, presenting clear limitations in chiral molecular simulation scenarios. To address this limitation, this paper proposes CE-ViSNet (Chiral Enhanced ViSNet), which constructs pseudoscalar chiral features based on the scalar triple product of nearest-neighbor atomic triplets and integrates them into the initial node embeddings without altering the core message-passing architecture. Evaluations on the MD17 and MD22 benchmark datasets demonstrate that CE-ViSNet achieves energy and force prediction accuracy comparable to the original ViSNet. Furthermore, in the chiral mirror-molecule recognition task on the QM9-OR dataset, its classification accuracy rises significantly from 48.89% to 81.20%. By retaining ViSNet's linear computational efficiency, this approach provides a lightweight machine learning force field for molecular dynamics simulations of chiral organic molecules.References
O. T. Unke, S. Chmiela, H. E. Sauceda, et al., "Machine learning force fields," Chemical Reviews, vol. 121, no. 16, pp. 10142–10186, 2021.
S. Batzner, A. Musaelian, L. Sun, et al., "E(3)-equivariant graph neural networks for data-efficient and accurate interatomic potentials," Nature Communications, vol. 13, no. 1, p. 2453, 2022.
Z. Yang, X. Wang, Y. Li, et al., "Efficient equivariant model for machine learning interatomic potentials," npj Computational Materials, vol. 11, no. 1, p. 49, 2025.
J. Gasteiger, F. Becker, and S. Günnemann, "GemNet: Universal directional graph neural networks for molecules," in Advances in Neural Information Processing Systems, vol. 34, pp. 6790–6802, 2021.
Y. Wang, T. Wang, S. Li, et al., "Enhancing geometric representations for molecules with equivariant vector-scalar interactive message passing," Nature Communications, vol. 15, no. 1, p. 313, 2024.
I. Batatia, D. P. Kovacs, G. Simm, et al., "MACE: Higher order equivariant message passing neural networks for fast and accurate force fields," in Advances in Neural Information Processing Systems, vol. 35, pp. 11423–11436, 2022.
A. Musaelian, S. Batzner, A. Johansson, et al., "Learning local equivariant representations for large-scale atomistic dynamics," Nature Communications, vol. 14, no. 1, p. 579, 2023.
K. T. Schütt, F. Arbabzadah, S. Chmiela, et al., "Quantum-chemical insights from deep tensor neural networks," Nature Communications, vol. 8, no. 1, p. 13890, 2017.
S. Chmiela, V. Vassilev-Galindo, O. T. Unke, et al., "Accurate global machine learning force fields for molecules with hundreds of atoms," Science Advances, vol. 9, no. 2, p. eadf0873, 2023.
R. Ramakrishnan, P. O. Dral, M. Rupp, et al., "Quantum chemistry structures and properties of 134 kilo molecules," Scientific Data, vol. 1, no. 1, p. 140022, 2014.
M. Sparta, C. Riplinger, and F. Neese, "Mechanism of olefin asymmetric hydrogenation catalyzed by iridium phosphino-oxazoline: a pair natural orbital coupled cluster study," Journal of Chemical Theory and Computation, vol. 10, no. 3, pp. 1099–1108, 2014.
Y. Zhou, H. Zhu, Y. Yuan, et al., "Machine learning classification of chirality and optical rotation using a simple one-hot encoded cartesian coordinate molecular representation," Journal of Chemical Information and Modeling, vol. 65, no. 9, p. 4281, 2025.
K. T. Schütt, H. E. Sauceda, P. J. Kindermans, et al., "SchNet—a deep learning architecture for molecules and materials," The Journal of Chemical Physics, vol. 148, no. 24, p. 241722, 2018.
J. Gasteiger, J. Groß, and S. Günnemann, "Directional message passing for molecular graphs," arXiv preprint arXiv:2003.03123, 2020.
K. Schütt, O. Unke, and M. Gastegger, "Equivariant message passing for the prediction of tensorial properties and molecular spectra," in International Conference on Machine Learning, PMLR, pp. 9377–9388, 2021.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Tao Luo (Author)
This work is licensed under a Creative Commons Attribution 4.0 International License.
