Learning 3D Anisotropic Noise Distributions Improves Molecular Force Field Modeling
This addresses the problem of inaccurate molecular force field modeling for computational chemistry and drug discovery, representing a novel method for a known bottleneck rather than a new paradigm.
The paper tackles the problem of oversimplified isotropic noise assumptions in 3D molecular denoising for force field modeling by proposing AniDS, an anisotropic variational autoencoder that generates atom-specific, full covariance matrices for noise distributions. The result shows average relative improvements of 8.9% and 6.2% in force prediction accuracy on MD17 and OC22 benchmarks compared to prior methods.
Coordinate denoising has emerged as a promising method for 3D molecular pretraining due to its theoretical connection to learning molecular force field. However, existing denoising methods rely on oversimplied molecular dynamics that assume atomic motions to be isotropic and homoscedastic. To address these limitations, we propose a novel denoising framework AniDS: Anisotropic Variational Autoencoder for 3D Molecular Denoising. AniDS introduces a structure-aware anisotropic noise generator that can produce atom-specific, full covariance matrices for Gaussian noise distributions to better reflect directional and structural variability in molecular systems. These covariances are derived from pairwise atomic interactions as anisotropic corrections to an isotropic base. Our design ensures that the resulting covariance matrices are symmetric, positive semi-definite, and SO(3)-equivariant, while providing greater capacity to model complex molecular dynamics. Extensive experiments show that AniDS outperforms prior isotropic and homoscedastic denoising models and other leading methods on the MD17 and OC22 benchmarks, achieving average relative improvements of 8.9% and 6.2% in force prediction accuracy. Our case study on a crystal and molecule structure shows that AniDS adaptively suppresses noise along the bonding direction, consistent with physicochemical principles. Our code is available at https://github.com/ZeroKnighting/AniDS.