Research map / Probabilistic and scientific modeling / Molecules and proteins
3DCS: Datasets and Benchmark for Evaluating Conformational Sensitivity in Molecular Representations
Xi Wang, Yang Zhang, Yingjia Zhang, Yejia Cai, Shenji Wan
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 |
|---|---|---|---|
| fine-tune with supervised contrastive loss on stereoisomers | UniMol · 3DCS chirality | ES-AUC: 0.622 -> 0.976 | Improved |
| fine-tune with supervised contrastive loss on stereoisomers | MolAE · 3DCS chirality | ES-AUC: 0.782 -> 0.971 | Improved |
| fine-tune with supervised contrastive loss on stereoisomers | FMG · 3DCS chirality | ES-AUC: 0.706 -> 0.999 | Improved |
| fine-tune with supervised contrastive loss on stereoisomers | MACE · 3DCS chirality | ES-AUC: 0.485 -> 0.887 | Improved |
| use cloze-style masked reconstruction instead of coordinate denoising pretraining | MolAE vs UniMol · 3DCS chirality (zero-shot) | ES-AUC: 0.622 -> 0.782 | Improved |
| add chiral-aware field-based encoding (FMG) versus equivariant encoders | FMG · 3DCS chirality (zero-shot) | ES-AUC: 0.622 -> 0.706 | Improved |
| add energy and force supervision with spherical harmonic inputs | MACE · 3DCS energy | Spearman: 0.026 -> 0.236 | Improved |
| use learned 3D encoder instead of E3FP fingerprint | UniMol · 3DCS geometry | Spearman: 0.406 -> 0.697 | Improved |
| use energy/force-trained encoder before RMD17 fine-tuning | GemNet · Revised MD17 | average energy MAE (kcal/mol): 3.941 -> 0.371 | Improved |
| add spectra (MolSpectra) to 3D denoising pretraining | MolSpectra vs UniMol · 3DCS geometry | Spearman: 0.697 -> 0.682 | No clear effect |
| use spectra-augmented pretraining on chirality | MolSpectra · 3DCS chirality (zero-shot) | ES-AUC: 0.622 -> 0.545 | Hurt |
Closest papers by meaning
- Learning Over Molecular Conformer Ensembles: Datasets and Benchmarks (ICLR 2024 · similarity 0.89)
- Beyond Atoms: Enhancing Molecular Pretrained Representations with 3D Space Modeling (ICML 2025 · similarity 0.88)
- Structure-Aware E(3)-Invariant Molecular Conformer Aggregation Networks (ICML 2024 · similarity 0.88)
- E(3)-equivariant models cannot learn chirality: Field-based molecular generation (ICLR 2025 · similarity 0.88)
- 3D Denoisers Are Good 2D Teachers: Molecular Pretraining via Denoising and Cross-Modal Distillation (AAAI 2025 · similarity 0.87)
- 3D Molecule Generation from Rigid Motifs via $\mathrm{SE}(3)$ Flows (NeurIPS 2026 · similarity 0.87)
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.
