Research map / Probabilistic and scientific modeling / Molecules and proteins
GeomMotif: A Benchmark for Arbitrary Geometric Preservation in Protein Generation
Pavel Strashnov, Andrey Shevtsov, Viacheslav Meshchaninov, Olga Kardymon, Dmitry Vetrov
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 structure modality to ESM3 (sequence and structure instead of sequence-only) | ESM3 · GeomMotif | SUN score (%): 3.5% -> 1% | Hurt |
| scale DPLM from 650M to 3B parameters | DPLM · GeomMotif | SUN score (%): 2.1% -> 2.7% | No clear effect |
| use broadly trained RFdiffusion instead of enzyme-specialised RFdiffusion2 | scaffolding model · GeomMotif | SUN score (%): 17.9% -> 37.8% | Improved |
| move from single-motif to paired-motif tasks | RFdiffusion · GeomMotif | SUN score (%): 62.4 -> 13.2 | Hurt |
| move from single-motif to paired-motif tasks | ESM3 (seq) · GeomMotif | SUN score (%): 6.8 -> 0.1 | Hurt |
| increase fragment count from 1 to 7 | RFdiffusion · GeomMotif single motifs | mean SUN score (%): 91.9% -> 38.7% | Hurt |
| increase fragment count from 1 to 7 | Genie2 · GeomMotif single motifs | mean SUN score (%): 76.7% -> 56.8% | Hurt |
| helical content above 70% instead of below 30% | Genie2 · GeomMotif single motifs | SUN score (%): 42.9% -> 84.6% | Improved |
| helical content above 70% instead of below 30% | FrameFlow · GeomMotif single motifs | SUN score (%): 14.0% -> 54.4% | Improved |
| exposed low-burial motifs instead of highly buried motifs with low contact ratio | Genie2 and RFdiffusion · GeomMotif single motifs | SUN score (%): below 30% -> 80.7% | Improved |
| purely helical paired motifs instead of mixed secondary structure | Genie2 · GeomMotif paired motifs | SUN score (%): 13.0% -> 46.5% | Improved |
| vary TM-score novelty/uniqueness threshold between 0.7 and 0.9 (ablation) | evaluation pipeline · GeomMotif | model ranking: 0.7 to 0.9 | No clear effect |
Closest papers by meaning
- Generating Highly Designable Proteins with Geometric Algebra Flow Matching (NeurIPS 2024 · similarity 0.87)
- ProteinBench: A Holistic Evaluation of Protein Foundation Models (ICLR 2025 · similarity 0.87)
- ProtComposer: Compositional Protein Structure Generation with 3D Ellipsoids (ICLR 2025 · similarity 0.87)
- Proteina: Scaling Flow-based Protein Structure Generative Models (ICLR 2025 · similarity 0.87)
- 3D Molecule Generation from Rigid Motifs via $\mathrm{SE}(3)$ Flows (NeurIPS 2026 · similarity 0.87)
- ProtPainter: Draw or Drag Protein via Topology-guided Diffusion (ICLR 2025 · 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.
