Denoise Pretraining on Nonequilibrium Molecules for Accurate and Transferable Neural Potentials
This work addresses the problem of limited data for molecular potential predictions in computational chemistry, offering a model-agnostic pretraining approach that is incremental but effective for enhancing GNN-based models.
The paper tackles the challenge of building accurate and transferable neural potentials for molecular systems using graph neural networks (GNNs), proposing denoise pretraining on nonequilibrium conformations to improve performance, with results showing significant accuracy improvements and remarkable transferability across diverse molecular systems.
Recent advances in equivariant graph neural networks (GNNs) have made deep learning amenable to developing fast surrogate models to expensive ab initio quantum mechanics (QM) approaches for molecular potential predictions. However, building accurate and transferable potential models using GNNs remains challenging, as the data is greatly limited by the expensive computational costs and level of theory of QM methods, especially for large and complex molecular systems. In this work, we propose denoise pretraining on nonequilibrium molecular conformations to achieve more accurate and transferable GNN potential predictions. Specifically, atomic coordinates of sampled nonequilibrium conformations are perturbed by random noises and GNNs are pretrained to denoise the perturbed molecular conformations which recovers the original coordinates. Rigorous experiments on multiple benchmarks reveal that pretraining significantly improves the accuracy of neural potentials. Furthermore, we show that the proposed pretraining approach is model-agnostic, as it improves the performance of different invariant and equivariant GNNs. Notably, our models pretrained on small molecules demonstrate remarkable transferability, improving performance when fine-tuned on diverse molecular systems, including different elements, charged molecules, biomolecules, and larger systems. These results highlight the potential for leveraging denoise pretraining approaches to build more generalizable neural potentials for complex molecular systems.