Research map / Graph learning / GNN architectures
Directional Message Passing on Molecular Graphs via Synthetic Coordinates
Johannes Gasteiger, Chandan Yeshwanth, Stephan Günnemann
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 |
|---|---|---|---|
| add PPR graph distance features (ablation) | DeeperGCN · ZINC | MAE: 0.317 -> 0.227 | Improved |
| add RDKit distance bounds features (ablation) | DeeperGCN · ZINC | MAE: 0.317 -> 0.264 | Improved |
| add line graph on top of distance (ablation) | DeeperGCN (Bounds+PPR) · ZINC | MAE: 0.228 -> 0.178 | Improved |
| add angle features on top of distance and line graph (ablation) | DeeperGCN (Bounds+PPR) · ZINC | MAE: 0.178 -> 0.142 | Improved |
| line graph and angles added to SMP with PPR (ablation) | SMP · ZINC | MAE: 0.151 -> 0.153 | No clear effect |
| combine bounds and PPR rather than bounds alone (ablation) | SMP full directional · ZINC | MAE: 0.113 -> 0.109 | Improved |
| use synthetic bounds/PPR coordinates instead of MMFF94/ETKDG conformers (ablation) | DeeperGCN · ZINC | MAE: 0.274 -> 0.142 (ETKDG vs Bounds+PPR) | Improved |
| add synthetic distances and angles (ablation) | DimeNet++ · coordinate-free QM9 U0 | MAE (meV): 41.9 -> 28.7 | Improved |
| add angles on top of distances (ablation) | DimeNet++ · coordinate-free QM9 U0 | MAE (meV): 32.1 -> 28.7 | Improved |
| convert DeeperGCN to line graph (directional MPNN) (ablation) | DeeperGCN · QM9 U0 | MAE (meV): 114 -> 233 (bounds) | Hurt |
| convert DeeperGCN to line graph (directional MPNN) (ablation) | DeeperGCN · ogbg-molhiv | ROC-AUC: 0.734 -> 0.697 (PPR) | Hurt |
| add PPR distance without line graph (ablation) | DeeperGCN · QM9 U0 / ogbg-molhiv | MAE / ROC-AUC: 106 -> 88; 0.728 -> 0.734 | Improved |
Closest papers by meaning
- Directional Message Passing for Molecular Graphs (ICLR 2020 · similarity 0.92)
- Neural Atoms: Propagating Long-range Interaction in Molecular Graphs through Efficient Communication Channel (ICLR 2024 · similarity 0.88)
- On the Expressive Power of Sparse Geometric MPNNs (ICLR 2025 · similarity 0.88)
- Lift Your Molecules: Molecular Graph Generation in Latent Euclidean Space (ICLR 2025 · similarity 0.87)
- On the Scalability of GNNs for Molecular Graphs (NeurIPS 2024 · similarity 0.86)
- Empowering Active Learning for 3D Molecular Graphs with Geometric Graph Isomorphism (NeurIPS 2024 · similarity 0.86)
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.
