Designing Protein Structures from Natural Language
We develop Prompteina, the first all-atom structure generation model that supports new and broader design objectives specified in natural language, analogous to text-to-image generation.
ProofPaper ↗
Key points
- A central challenge in de novo protein design is generating proteins with desired structural and functional properties.
- To train Prompteina, we construct Prompteina-Data, a paired text--protein structure dataset of 1.8M structures and 50M+ captions describing global and residue-specific structural and functional properties.
- We adapt Proteina-Complexa, a pretrained structure generation model, through text-conditioned pretraining and finetuning that aligns residues with their corresponding text.
- These results establish natural language as an effective and powerful conditioning interface for protein structure generation.
Sources (1)
- [1]Designing Protein Structures from Natural LanguagearXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 9, 05:51 PM
We develop Prompteina, the first all-atom structure generation model that supports new and broader design objectives specified in natural language, analogous to text-to-image generation.
A central challenge in de novo protein design is generating proteins with desired structural and functional properties.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 9, 2026A Critical Look at Power Sampling when Test Time Sharpening LLMs
- Oct 9, 2026Mix, Don't Tune: Bilingual Pre-Training Outperforms Hyperparameter Search in Data-Constrained Settings
- Oct 9, 2026Minions: A Dynamic and Distributed Virtual GPU Cloud Built on Idle Apple Silicon Devices
- Oct 9, 2026Rethinking Relative Policy Differences Transfer of On-Policy Distillation
- Oct 9, 2026Latent World Action Model for GUI Agents
- Oct 9, 2026InstantFusion: A Shared Latent Interface for Heterogeneous Image Generators