Research map / Graph learning / Graph self-supervised learning
Molformer: Motif-Based Transformer on 3D Heterogeneous Molecular Graphs
Fang Wu, Dragomir Radev, Stan Z. Li
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 motif and atom nodes as heterogeneous molecular graphs (ablation) | Molformer · QM7 | MAE: 132.2 -> 43.2 | Improved |
| add motif and atom nodes as heterogeneous molecular graphs (ablation) | Molformer · QM8 | MAE: 0.0205 -> 0.0095 | Improved |
| add motif and atom nodes as heterogeneous molecular graphs (ablation) | Molformer · PDBbind | RMSD: 1.925 -> 1.386 | Improved |
| use separate heterogeneous motif nodes instead of multi-level feature fusion (ablation) | Molformer · QM7 | MAE: 89.7 -> 43.2 | Improved |
| use separate heterogeneous motif nodes instead of multi-level feature fusion (ablation) | Molformer · QM8 | MAE: 0.0154 -> 0.0095 | Improved |
| use separate heterogeneous motif nodes instead of multi-level feature fusion (ablation) | Molformer · PDBbind | RMSD: 1.427 -> 1.386 | Improved |
| add HSA and HMG instead of passing 3D coordinates directly as atom input features (ablation) | Molformer · QM7 | MAE: 132.2 -> 46.5 | Improved |
| add HSA and HMG instead of passing 3D coordinates directly as atom input features (ablation) | Molformer · QM8 | MAE: 0.0205 -> 0.0097 | Improved |
| add AFPS molecular readout (ablation) | Molformer · QM7 | MAE: 46.5 -> 43.2 | Improved |
| add AFPS molecular readout (ablation) | Molformer · PDBbind | RMSD: 1.441 -> 1.386 | Improved |
| add multi-scale distance-constrained attention (ablation) | Molformer · QM7 | MAE: 50.1 -> 46.5 | Improved |
| add multi-scale distance-constrained attention (ablation) | Molformer · QM8 | MAE | No clear effect |
Later papers that compared with this method
2 later accepted papers report MolFormer as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Monroe: A Molecular Foundation Model for In-context Probabilistic Inference (NeurIPS 2026 · better on 4 of 4 shared results)
- CL-MFAP: A Contrastive Learning-Based Multimodal Foundation Model for Molecular Property Prediction and Antibiotic Screening (ICLR 2025 · better on 3 of 6 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- 3DMolFormer: A Dual-channel Framework for Structure-based Drug Discovery (ICLR 2025 · similarity 0.88)
- DMol: A Highly Efficient and Chemical Motif-Preserving Molecule Generation Platform (NeurIPS 2025 · similarity 0.88)
- Empowering Active Learning for 3D Molecular Graphs with Geometric Graph Isomorphism (NeurIPS 2024 · similarity 0.87)
- Beyond Atoms: Enhancing Molecular Pretrained Representations with 3D Space Modeling (ICML 2025 · similarity 0.87)
- Representing Molecules as Random Walks Over Interpretable Grammars (ICML 2024 · similarity 0.86)
- Learning Over Molecular Conformer Ensembles: Datasets and Benchmarks (ICLR 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.
