Research map / Graph learning / Graph self-supervised learning
Learning 3D Representations of Molecular Chirality with Invariance to Bond Rotations
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Encode coupled torsions with learned phase shifts and internal-bond rotation invariance | ChIRo · l / d classification | accuracy (%): 79.3 ± 0.4 vs 65.5 ± 2.4 | Helped |
| Encode coupled torsions with learned phase shifts and internal-bond rotation invariance | ChIRo · Enantiomer ranking | accuracy (%): 72.0 ± 0.5 vs 68.6 ± 0.3 | Helped |
| Use one pre-selected conformer rather than conformer sampling (ablation) | ChIRo · l / d classification | accuracy (%): 79.3 ± 0.4 -> 79.1 ± 0.5 | No clear effect |
| Use one pre-selected conformer rather than conformer sampling (ablation) | ChIRo · Enantiomer ranking | accuracy (%): 72.0 ± 0.5 -> 71.9 ± 0.3 | No clear effect |
| Add stereochemical tags to the full model (ablation) | ChIRo · l / d classification | accuracy (%): 79.3 ± 0.4 -> 77.7 ± 0.8 | Hurt |
| Add stereochemical tags to the full model (ablation) | ChIRo · Enantiomer ranking | accuracy (%): 72.0 ± 0.5 -> 71.3 ± 0.7 | Hurt |
| Mask internal coordinates while retaining conformer embeddings at readout (ablation) | ChIRo with tags · l / d classification | accuracy (%): 77.7 ± 0.8 -> 76.2 ± 0.6 | Hurt |
| Mask internal coordinates while retaining conformer embeddings at readout (ablation) | ChIRo with tags · Enantiomer ranking | accuracy (%): 71.3 ± 0.7 -> 68.4 ± 0.9 | Hurt |
| Omit conformer embeddings from readout when coordinates are masked (ablation) | ChIRo with tags · l / d classification | accuracy (%): 76.2 ± 0.6 -> 70.0 ± 0.6 | Hurt |
| Omit conformer embeddings from readout when coordinates are masked (ablation) | ChIRo with tags · Enantiomer ranking | accuracy (%): 68.4 ± 0.9 -> 63.7 ± 0.2 | Hurt |
| Remove both stereochemical tags and internal-coordinate information (ablation) | ChIRo · l / d classification | accuracy (%): 79.3 ± 0.4 -> 50.0 ± 0.0 | Hurt |
| Remove both stereochemical tags and internal-coordinate information (ablation) | ChIRo · Enantiomer ranking | accuracy (%): 72.0 ± 0.5 -> 49.6 ± 0.3 | Hurt |
Closest papers by meaning
- E(3)-equivariant models cannot learn chirality: Field-based molecular generation (ICLR 2025 · similarity 0.91)
- Learning Over Molecular Conformer Ensembles: Datasets and Benchmarks (ICLR 2024 · similarity 0.90)
- A Plug-and-Play Quaternion Message-Passing Module for Molecular Conformation Representation (AAAI 2024 · similarity 0.89)
- Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks (ICML 2024 · similarity 0.87)
- Empowering Active Learning for 3D Molecular Graphs with Geometric Graph Isomorphism (NeurIPS 2024 · similarity 0.86)
- Predicting directional flexibility in proteins (NeurIPS 2026 · 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.