GemNet: Universal Directional Graph Neural Networks for Molecules

Abstract

Effectively predicting molecular interactions has the potential to accelerate molecular dynamics by multiple orders of magnitude and thus revolutionize chemical simulations. Graph neural networks (GNNs) have recently shown great successes for this task, overtaking classical methods based on fixed molecular kernels. However, they still appear very limited from a theoretical perspective, since regular GNNs cannot distinguish certain types of graphs. In this work we close this gap between theory and practice. We show that GNNs with spherical representations are indeed universal approximators for predictions that are invariant to translation, and equivariant to permutation and rotation. We then discretize such GNNs via directed edge embeddings and two-hop message passing, and incorporate multiple structural improvements to arrive at the geometric message passing neural network (GemNet). We demonstrate the benefits of the proposed changes in multiple ablation studies. GemNet outperforms previous models on the COLL, MD17, and OC20 datasets by 34%, 41%, and 20%, respectively, and performs especially well on the most challenging molecules. Our implementation is available online.

arXiv commentPublished as a conference paper at NeurIPS 2021. Author name changed from Johannes Klicpera to Johannes Gasteiger

Cited by in corpus (15)

• E(3)-Equivariant Graph Neural Networks for Data-Efficient and Accurate Interatomic Potentials
• Predicting binding motifs of complex adsorbates using machine learning with a physics-inspired graph representation
• Force-Field-Enhanced Neural Network Interactions: from Local Equivariant Embedding to Atom-in-Molecule properties and long-range effects
• Artificial Intelligence in Materials Science and Engineering: Current Landscape, Key Challenges, and Future Trajectorie
• Automated 3D Pre-Training for Molecular Property Prediction
• Accelerating the prediction of inorganic surfaces with machine learning interatomic potentials
• Data efficiency and extrapolation trends in neural network interatomic potentials
• Boosting Heterogeneous Catalyst Discovery by Structurally Constrained Deep Learning Models
• Machine learning and invariant theory
• Towards Foundation Models for Materials Science: The Open MatSci ML Toolkit
• MuGSI: Distilling GNNs with Multi-Granularity Structural Information for Graph Classification
• Improving the Expressiveness of $K$-hop Message-Passing GNNs by Injecting Contextualized Substructure Information
• Scaling Graph Neural Networks to Large Proteins
• Complete Neural Networks for Complete Euclidean Graphs
• Knowledge Distillation of a Protein Language Model Yields a Foundational Implicit Solvent Model