Research map / Probabilistic and scientific modeling / Molecules and proteins
EPO: Diverse and Realistic Protein Ensemble Generation via Energy Preference Optimization
Yuancheng Sun, Yuxuan Ren, Zhaoming Chen, Xu Han, Kang Liu, Qiwei Ye
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 |
|---|---|---|---|
| use listwise instead of pairwise energy preference loss (ablation) | EPO · Fast-Folding Rg JSD (mean/median) | JSD: 2.68/2.54 -> 2.36/2.28 | Improved |
| use listwise instead of pairwise energy preference loss (ablation) | EPO · Fast-Folding RMSF JSD | JSD: 0.39/0.40 -> 0.32/0.30 | Improved |
| use listwise instead of pairwise energy preference loss (ablation) | EPO · Tetrapeptides | JSD TICA-0,1: 0.318 -> 0.311 | Improved |
| use listwise instead of pairwise energy preference loss (ablation) | EPO · ATLAS | %PC-sim>0.5: 9 -> 13 | Improved |
| use SDE instead of ODE sampling during online optimization (ablation) | EPO · Tetrapeptides | JSD chi_all: 0.132 -> 0.107 | Improved |
| use SDE instead of ODE sampling during online optimization (ablation) | EPO · Tetrapeptides TICA-0,1 | JSD: 0.337 -> 0.311 | Improved |
| apply EPO online energy preference fine-tuning to pretrained MDGen | MDGen · ATLAS | Exposed residue J: 0.29 -> 0.44 | Improved |
| apply EPO fine-tuning to pretrained MDGen | MDGen · ATLAS | Weak contacts J: 0.51 -> 0.45 | Hurt |
| apply EPO fine-tuning to pretrained MDGen | MDGen · ATLAS | Root W2: 2.69/3.47 -> 2.79/3.43 | Hurt |
| apply EPO fine-tuning to pretrained MDGen | MDGen · Tetrapeptides | JSD chi_all: 0.109 -> 0.107 | Improved |
| use a very small beta (ablation) | EPO · Tetrapeptides | distribution quality | Hurt |
| use too few denoising steps in SDE sampling (ablation) | EPO · Tetrapeptides | training stability | Hurt |
Closest papers by meaning
- Aligning Protein Conformation Ensemble Generation with Physical Feedback (ICML 2025 · similarity 0.91)
- AlphaFold Meets Flow Matching for Generating Protein Ensembles (ICML 2024 · similarity 0.87)
- JAMUN: Bridging Smoothed Molecular Dynamics and Score-Based Learning for Conformational Ensemble Generation (NeurIPS 2025 · similarity 0.87)
- Action-Minimization Meets Generative Modeling: Efficient Transition Path Sampling with the Onsager-Machlup Functional (ICML 2025 · similarity 0.87)
- TEMPO: Temporal Multi-scale Autoregressive Generation of Protein Conformational Ensembles (NeurIPS 2025 · similarity 0.86)
- Str2Str: A Score-based Framework for Zero-shot Protein Conformation Sampling (ICLR 2024 · 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.
