Research map / Probabilistic and scientific modeling / Molecules and proteins
PepCompass: Navigating Peptide Embedding Spaces Using Riemannian Geometry
Marcin Możejko, Adam Bielecki, Jurand Prądzyński, Hyun-Su Lee, Antoni Janowski, Michal Kmicikiewicz, Paulina Szymczak, Karol Jurasz, Marcin Traskowski, Michał Kucharczyk, Marcelo Der Torossian Torres, Cesar de la Fuente-Nunez and 1 more
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 property potential term to geodesic search (lambda=0.01) (ablation) | PoGS · PoGS, 300 prototype pairs | seeds per path: 18.90 -> 22.70 | Improved |
| add property potential term to geodesic search (lambda=0.01) (ablation) | PoGS · PoGS, 300 prototype pairs | wells per path: 6.40 -> 12.60 | Improved |
| use geodesic search instead of straight latent interpolation | PoGS · PoGS, 300 prototype pairs | seeds per path: 11.70 -> 22.70 | Improved |
| potential term increases ambient path length vs no-potential geodesic (ablation) | PoGS · PoGS, 300 prototype pairs | ambient length: 1240.03 -> 1432.55 | Hurt |
| replace Euclidean random walk with SORBES-SE Riemannian walk (with mutations) (ablation) | LE-BO · AMP optimization, KY14 | min log2 MIC: 0.65 -> 0.50 | Improved |
| add MUTANG mutation enumeration to SORBES-SE walk (ablation) | LE-BO · AMP optimization, KY14 | min log2 MIC: 1.37 -> 0.50 | Improved |
| add Riemannian random walk to mutation-only enumeration (ablation) | LE-BO · AMP optimization, KK16 | min log2 MIC: 1.82 -> 0.50 | Improved |
| LE-BO vs Random Mutation baseline | LE-BO · AMP optimization, mammuthusin-3 | min log2 MIC: 2.12 -> 0.50 | Improved |
| LE-BO vs AdaLead on FL14 | LE-BO · AMP optimization, FL14 | min log2 MIC: 0.51 -> 0.60 | Hurt |
| LE-BO optimization of PoGS seeds, wet-lab success at 32 ug/ml | PepCompass · In-vitro, 29 peptides, 19 strains | success rate: 100% | Improved |
| Euclidean walk with no mutation vs SORBES-SE walk with no mutation (ablation) | LE-BO · AMP optimization, six prototypes | min log2 MIC: 1.37 vs 1.37 (KY14) | No clear effect |
Closest papers by meaning
- Full-Atom Peptide Design with Geometric Latent Diffusion (NeurIPS 2024 · similarity 0.88)
- Full-Atom Peptide Design based on Multi-modal Flow Matching (ICML 2024 · similarity 0.85)
- PPFLOW: Target-Aware Peptide Design with Torsional Flow Matching (ICML 2024 · similarity 0.85)
- SE(3)-Stochastic Flow Matching for Protein Backbone Generation (ICLR 2024 · similarity 0.84)
- AlphaFold Meets Flow Matching for Generating Protein Ensembles (ICML 2024 · similarity 0.84)
- Swallowing the Bitter Pill: Simplified Scalable Conformer Generation (ICML 2024 · similarity 0.84)
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.
