Research map / Probabilistic and scientific modeling / Molecules and proteins
Symmetry-Informed Geometric Representation for Molecules, Proteins, and Crystalline Materials
Shengchao Liu, weitao Du, Yanjing Li, Zhuoxinran Li, Zhiling Zheng, Chenru Duan, Zhi-Ming Ma, Omar M. Yaghi, Anima Anandkumar, Christian Borgs, Jennifer T Chayes, Hongyu Guo and 1 more
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use 3D conformation instead of 2D graph features | GNN models · QM9 U0 | MAE: 411.16 -> 13.27 | Improved |
| add data normalization (scale energy by mean force) (ablation) | NequIP/Allegro-style models · MD17 and rMD17 | force MAE | Improved |
| use gathered instead of expanded periodic data augmentation (ablation) | geometric models · MatBench | MAE | Improved |
| use gathered instead of expanded DA (ablation) | EGNN · MatBench | MAE | Hurt |
| supervised pretraining on PCQM4Mv2 (ablation) | SchNet · QM9 gap | MAE: 44.13 -> 40.31 | Improved |
| supervised pretraining on PCQM4Mv2 (ablation) | SchNet · QM9 R2 | MAE: 0.133 -> 0.308 | Hurt |
| MoleculeSDE (VE) pretraining (ablation) | SchNet · QM9 U0 | MAE: 13.27 -> 10.95 | Improved |
| GeoSSL-DDM pretraining (ablation) | SchNet · QM9 U0 | MAE: 13.27 -> 10.96 | Improved |
| MoleculeSDE (VE) pretraining (ablation) | SchNet · QM9 gap | MAE: 44.13 -> 41.84 | Improved |
| Type Prediction pretraining (ablation) | SchNet · QM9 gap | MAE: 44.13 -> 45.38 | Hurt |
| Angle Prediction pretraining (ablation) | SchNet · QM9 gap | MAE: 44.13 -> 48.45 | Hurt |
| use PaiNN instead of GemNet-T/SphereNet | PaiNN · QM9 | training cost | Improved |
Closest papers by meaning
- 3DCS: Datasets and Benchmark for Evaluating Conformational Sensitivity in Molecular Representations (ICLR 2026 · similarity 0.89)
- GeoMFormer: A General Architecture for Geometric Molecular Representation Learning (ICML 2024 · similarity 0.88)
- HAGO-Net: Hierarchical Geometric Message Passing for Molecular Representation Learning (AAAI 2024 · similarity 0.88)
- Local-Global Associative Frames for Symmetry-Preserving Crystal Structure Modeling (NeurIPS 2025 · similarity 0.87)
- Conformal Crystal Graph Transformer with Robust Encoding of Periodic Invariance (AAAI 2024 · similarity 0.87)
- Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks (ICML 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
