DeepChem Equivariant: SE(3)-Equivariant Support in an Open-Source Molecular Machine Learning Library

2025-10-19

AI Summary

Existing SE(3)-equivariant neural network libraries (e.g., E3NN, SE(3)-Transformer) require substantial expertise in deep learning or group representation theory and lack end-to-end training pipelines, severely limiting adoption by non-specialist researchers in molecular modeling. To address this, we present the first systematic integration of SE(3)-equivariance into DeepChem, incorporating state-of-the-art equivariant architectures—including SE(3)-Transformer and Tensor Field Networks—within a unified framework. The implementation supports key tasks such as molecular property prediction and protein structure modeling. We provide production-ready model interfaces, standardized training and evaluation workflows, comprehensive documentation, and rigorous unit tests. This work significantly lowers the technical barrier for 3D geometrically aware modeling, enabling broader accessibility of SE(3)-equivariant methods among computational chemists and structural biologists without specialized mathematical background.

Technology Category

• Machine Learning: Deep Neural Architectures and Foundation Models
• Computer Vision: 3D Computer Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Neural networks that incorporate geometric relationships respecting SE(3) group transformations (e.g. rotations and translations) are increasingly important in molecular applications, such as molecular property prediction, protein structure modeling, and materials design. These models, known as SE(3)-equivariant neural networks, ensure outputs transform predictably with input coordinate changes by explicitly encoding spatial atomic positions. Although libraries such as E3NN [4] and SE(3)-TRANSFORMER [3 ] offer powerful implementations, they often require substantial deep learning or mathematical prior knowledge and lack complete training pipelines. We extend DEEPCHEM [ 13] with support for ready-to-use equivariant models, enabling scientists with minimal deep learning background to build, train, and evaluate models, such as SE(3)-Transformer and Tensor Field Networks. Our implementation includes equivariant models, complete training pipelines, and a toolkit of equivariant utilities, supported with comprehensive tests and documentation, to facilitate both application and further development of SE(3)-equivariant models.

Problem

• Enabling easy use of SE(3)-equivariant neural networks
• Providing complete training pipelines for molecular applications
• Facilitating model building for non-deep learning experts

Innovation

• Extends DeepChem with SE(3)-equivariant model support
• Provides complete training pipelines for molecular applications
• Offers equivariant utilities with tests and documentation