Score: 0

Scalable Machine Learning Force Fields for Macromolecular Systems Through Long-Range Aware Message Passing

Published: January 7, 2026 | arXiv ID: 2601.03774v1

By: Chu Wang , Lin Huang , Xinran Wei and more

Potential Business Impact:

Lets computers see how big things move.

Business Areas:
Quantum Computing Science and Engineering

Machine learning force fields (MLFFs) have revolutionized molecular simulations by providing quantum mechanical accuracy at the speed of molecular mechanical computations. However, a fundamental reliance of these models on fixed-cutoff architectures limits their applicability to macromolecular systems where long-range interactions dominate. We demonstrate that this locality constraint causes force prediction errors to scale monotonically with system size, revealing a critical architectural bottleneck. To overcome this, we establish the systematically designed MolLR25 ({Mol}ecules with {L}ong-{R}ange effect) benchmark up to 1200 atoms, generated using high-fidelity DFT, and introduce E2Former-LSR, an equivariant transformer that explicitly integrates long-range attention blocks. E2Former-LSR exhibits stable error scaling, achieves superior fidelity in capturing non-covalent decay, and maintains precision on complex protein conformations. Crucially, its efficient design provides up to 30% speedup compared to purely local models. This work validates the necessity of non-local architectures for generalizable MLFFs, enabling high-fidelity molecular dynamics for large-scale chemical and biological systems.

Country of Origin
πŸ‡­πŸ‡° Hong Kong

Page Count
31 pages

Category
Physics:
Chemical Physics