Research map / Probabilistic and scientific modeling / Molecules and proteins
Intrinsic-Extrinsic Convolution and Pooling for Learning on 3D Protein Structures
Pedro Hermosilla, Marco Schäfer, Matej Lang, Gloria Fackelmann, Pere-Pau Vázquez, Barbora Kozlikova, Michael Krone, Tobias Ritschel, Timo Ropinski
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 |
|---|---|---|---|
| Combine covalent-bond and hydrogen-bond graph distances with Euclidean distance in convolution kernels (ablation) | IEConv · SCOPe 1.75 Fold | mean accuracy: 30.1\,\% -> 45.0\,\% | Improved |
| Combine covalent-bond and hydrogen-bond graph distances with Euclidean distance in convolution kernels (ablation) | IEConv · Enzyme reaction classification | mean accuracy: 85.0\,\% -> 87.2\,\% | Improved |
| Add intrinsic covalent-bond distance to hydrogen-bond distance and Euclidean distance (ablation) | IEConv · SCOPe 1.75 Fold | mean accuracy: 40.8\,\% -> 45.0\,\% | Improved |
| Use Euclidean distance in addition to both intrinsic distances (ablation) | IEConv · SCOPe 1.75 Fold | mean accuracy: 43.5\,\% -> 45.0\,\% | Improved |
| Use Euclidean-distance neighborhoods rather than covalent-graph neighborhoods (ablation) | IEConv · SCOPe 1.75 Fold | mean accuracy: 27.2\,\% -> 45.0\,\% | Improved |
| Use Euclidean-distance neighborhoods rather than hydrogen-bond-graph neighborhoods (ablation) | IEConv · Enzyme reaction classification | mean accuracy: 56.9 \% -> 87.2\,\% | Improved |
| Add protein-specific hierarchical pooling (ablation) | IEConv · SCOPe 1.75 Fold | mean accuracy: 37.1\,\% -> 45.0\,\% | Improved |
| Use protein-specific pooling rather than grid pooling (ablation) | IEConv · Enzyme reaction classification | mean accuracy: 86.1 \% -> 87.2\,\% | Hurt |
| Use protein-specific pooling rather than EdgePool (ablation) | IEConv · Enzyme reaction classification | mean accuracy: 86.9 \% -> 87.2\,\% | Improved |
| Represent proteins at atomic rather than amino-acid resolution (ablation) | IEConv · SCOPe 1.75 Fold | mean accuracy: 39.6\,\% -> 45.0\,\% | Improved |
| Use protein-specific pooling rather than EdgePool (ablation) | IEConv · SCOPe 1.75 Family | mean accuracy: 99.0\,\% -> 98.9\,\% | Hurt |
| Use protein-specific pooling rather than amino-acid-level representation (ablation) | IEConv · SCOPe 1.75 Family | mean accuracy: 99.1\,\% -> 98.9\,\% | Hurt |
Closest papers by meaning
- Evaluating Representation Learning on the Protein Structure Universe (ICLR 2024 · similarity 0.88)
- AtomSurf: Surface Representation for Learning on Protein Structures (ICLR 2025 · similarity 0.87)
- Towards Multiscale Graph-based Protein Learning with Geometric Secondary Structural Motifs (NeurIPS 2025 · similarity 0.87)
- Clustering for Protein Representation Learning (CVPR 2024 · similarity 0.86)
- Protein 3D Graph Structure Learning for Robust Structure-Based Protein Property Prediction (AAAI 2024 · similarity 0.86)
- Learning Complete Protein Representation by Dynamically Coupling of Sequence and Structure (NeurIPS 2024 · 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.
