PROTEAN is a standardized benchmark for text-guided protein structural editing, showing that frontier LLMs achieve 46% geometric compliance but only 7% end-to-end biophysical success, with backend scope leakage and refolding failure as primary bottlenecks.
and Li, Yanjing and Corley, Nathaniel and Kim, Paul T
8 Pith papers cite this work, alongside 42 external citations. Polarity classification is still indexing.
8
Pith papers citing it
42
external citations · external index
citation-role summary
background 3
citation-polarity summary
years
2026 8roles
background 3representative citing papers
ProtDBench is a new evaluation benchmark that standardizes protein binder design assessment, reveals verifier-dependent bias in structure predictors, and compares generative methods under fixed 24-hour and diversity-aware criteria.
Diffusion over DCT spectral volumes of Cα displacements yields fast, temperature-conditioned protein trajectories with RMSF Pearson r of 0.844 on held-out mdCATH.
ReproBAIT finds that current coding agents can reproduce select biological AI tools from public descriptions at low cost, with performance matching or exceeding the original on two of three tools tested.
ELMS couples LLM-based critiquing, motif-locked edits, and Monte Carlo tree search so folded-structure feedback is reused to steer scaffolding, nearly doubling task success at a fixed design budget.
Context-parallel inference for RFdiffusion 3 enables end-to-end all-atom design of large symmetric protein nanoparticles on multi-GPU hardware without retraining.
A ~3.3M-parameter dense PairMixer inverse folder trails LigandMPNN on interface recovery while propagating ligand signal to distal residues where local graphs lose it.
citing papers explorer
-
PROTEAN: Beyond De Novo Generation - Can LLM Agents Perform Text-Guided Protein Structure Editing?
PROTEAN is a standardized benchmark for text-guided protein structural editing, showing that frontier LLMs achieve 46% geometric compliance but only 7% end-to-end biophysical success, with backend scope leakage and refolding failure as primary bottlenecks.
-
ProtDBench: A Unified Benchmark of Protein Binder Design and Evaluation
ProtDBench is a new evaluation benchmark that standardizes protein binder design assessment, reveals verifier-dependent bias in structure predictors, and compares generative methods under fixed 24-hour and diversity-aware criteria.
-
Spectral Diffusion for Protein Dynamics
Diffusion over DCT spectral volumes of Cα displacements yields fast, temperature-conditioned protein trajectories with RMSF Pearson r of 0.844 on held-out mdCATH.
-
ReproBAIT: Can coding agents build their own biological AI tools?
ReproBAIT finds that current coding agents can reproduce select biological AI tools from public descriptions at low cost, with performance matching or exceeding the original on two of three tools tested.
-
Evaluator-in-the-Loop Monte Carlo Tree Search via LLM Agents for Motif Scaffolding in Protein Design
ELMS couples LLM-based critiquing, motif-locked edits, and Monte Carlo tree search so folded-structure feedback is reused to steer scaffolding, nearly doubling task success at a fixed design budget.
-
Design-CP: Context Parallelism for Design of Protein Nanoparticles
Context-parallel inference for RFdiffusion 3 enables end-to-end all-atom design of large symmetric protein nanoparticles on multi-GPU hardware without retraining.
-
UMA-Inverse: Ligand-Conditioned Protein Inverse Folding with a Distogram-Supervised Dense Pair Encoder
A ~3.3M-parameter dense PairMixer inverse folder trails LigandMPNN on interface recovery while propagating ligand signal to distal residues where local graphs lose it.
- Emyx: Fast and efficient all-atom protein generation