Research map / Label-efficient and robust learning / OOD and anomaly detection
Learning Substructure Invariance for Out-of-Distribution Molecular Representations
Nianzu Yang, Kaipeng Zeng, Qitian Wu, Xiaosong Jia, Junchi Yan
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 attention-based substructure architecture with ERM loss (ablation) | GIN + ERM · EC50-assay | ROC-AUC (%): 69.35 -> 71.44 | Improved |
| add attention-based substructure architecture with ERM loss (ablation) | GIN + ERM · EC50-size | ROC-AUC (%): 60.94 -> 64.23 | Improved |
| add new learning objective on top of architecture (full model) (ablation) | Our architecture + ERM loss · EC50-assay | ROC-AUC (%): 71.44 -> 73.25 | Improved |
| add new learning objective on top of architecture (full model) (ablation) | Our architecture + ERM loss · EC50-size | ROC-AUC (%): 64.23 -> 65.09 | Improved |
| add substructure architecture to GIN + new objective (ablation) | GIN + new learning objective · EC50-assay | ROC-AUC (%): 72.07 -> 73.25 | Improved |
| replace given environment labels with inferred environments (ablation) | Our model · EC50-assay | ROC-AUC (%): 71.94 -> 73.25 | Improved |
| replace given environment labels with inferred environments (ablation) | Our model · EC50-size | ROC-AUC (%): 63.38 -> 65.09 | Improved |
| use inferred environment labels instead of given ones (ablation) | DANN · EC50-assay | ROC-AUC (%): 66.97 -> 68.83 | Improved |
| use inferred environment labels instead of given ones (ablation) | DANN · EC50-size | ROC-AUC (%): 61.11 -> 62.56 | Improved |
| adapt backbone with our method | GCN · BACE | ROC-AUC (%): 80.01 -> 84.33 | Improved |
| adapt backbone with our method | GIN · HIV | ROC-AUC (%): 76.58 -> 78.31 | Improved |
| add virtual node to GCN | GCN · BACE | ROC-AUC (%): 80.01 -> 77.51 | Hurt |
Closest papers by meaning
- Learning Over Molecular Conformer Ensembles: Datasets and Benchmarks (ICLR 2024 · similarity 0.87)
- Multimodal Molecular Pretraining via Modality Blending (ICLR 2024 · similarity 0.86)
- Omni-Mol: Multitask Molecular Model for Any-to-any Modalities (NeurIPS 2025 · similarity 0.86)
- Mol-AE: Auto-Encoder Based Molecular Representation Learning With 3D Cloze Test Objective (ICML 2024 · similarity 0.86)
- MORE: Molecule Pretraining with Multi-Level Pretext Task (AAAI 2025 · similarity 0.86)
- Atomas: Hierarchical Adaptive Alignment on Molecule-Text for Unified Molecule Understanding and Generation (ICLR 2025 · similarity 0.85)
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.
