Research map / Probabilistic and scientific modeling / Molecules and proteins
Physically Valid Biomolecular Interaction Modeling with Gauss-Seidel Projection
Siyuan Chen, Minghao Guo, Caoliwen Wang, Anka He Chen, Yikun Zhang, Jingjing Chai, Yin Yang, Wojciech Matusik, Peter Yichen Chen
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 Gauss-Seidel projection at inference only (ablation) | 2-step Boltz-1 · CASP15 (2-step Boltz-1) | Physical validity: 0.6327 -> 1.00 | Improved |
| add Gauss-Seidel projection at inference only (ablation) | 2-step Boltz-1 · CASP15 (2-step Boltz-1) | Complex LDDT: 0.5920 -> 0.5889 | Hurt |
| add gradient descent guidance at inference (ablation) | 2-step Boltz-1 · CASP15 (2-step Boltz-1) | Physical validity: 0.6327 -> 0.7345 | Improved |
| add gradient descent guidance at inference (ablation) | 2-step Boltz-1 · CASP15 (2-step Boltz-1) | Complex LDDT: 0.5920 -> 0.5942 | No clear effect |
| finetune with differentiable GS projection (ablation) | 2-step Boltz-1 + GS · CASP15 (2-step Boltz-1) | Complex LDDT: 0.5889 -> 0.6321 | Improved |
| reduce denoising steps from 200 to 2 without guidance (ablation) | Boltz-1 · CASP15 (2-step Boltz-1) | Complex LDDT: 0.6406 -> 0.5920 | Hurt |
| use Gauss-Seidel projection instead of gradient descent guidance | physics potential minimization · PDB 8X51 noised | iterations to convergence | Improved |
| use small alpha (<=1e-6) in projection (ablation) | GS projection · PDB 7WUY | convergence to valid state | Improved |
| 2-step finetuned GS model vs Protenix-Mini 5-step | Ours · Test | Complex LDDT: 0.73 -> 0.79 | Improved |
| apply GS projection to Boltz-2 outputs | Boltz-2 · CASP15 | Physical validity | Improved |
Closest papers by meaning
- Few-Step Cofolding with All-Atom Flow Maps (NeurIPS 2026 · similarity 0.86)
- Consistent Sampling and Simulation: Molecular Dynamics with Energy-Based Diffusion Models (NeurIPS 2025 · similarity 0.86)
- Unlocking hidden biomolecular conformational landscapes in diffusion models at inference time (NeurIPS 2025 · similarity 0.85)
- CausalBind: Causal Modeling and Learning for Protein-Molecule Virtual Screening (NeurIPS 2026 · similarity 0.84)
- ShEPhERD: Diffusing shape, electrostatics, and pharmacophores for bioisosteric drug design (ICLR 2025 · similarity 0.84)
- Boltzmann-Aligned Inverse Folding Model as a Predictor of Mutational Effects on Protein-Protein Interactions (ICLR 2025 · 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.
