Aligning Proteins and Language: A Foundation Model for Protein Retrieval

Qifeng Wu, Zhengzhe Liu, Han Zhu, Yizhou Zhao, Daisuke Kihara, Min Xu, [3pt] Carnegie Mellon University, Purdue University · 2025

Read Aligning Proteins and Language: A Foundation Model for Protein Retrieval by Qifeng Wu, Zhengzhe Liu, Han Zhu, Yizhou Zhao, Daisuke Kihara, Min Xu, [3pt] Carnegie Mellon University, Purdue University (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-08023.txt.

Description

Qifeng Wu and five co-authors build a search engine that finds proteins by shape and meaning together. Labs increasingly solve protein structures without knowing what the proteins do. The team borrows the recipe of vision-language models: train one network on structures and another on written descriptions, pulling matching pairs together and pushing others apart.

They assembled about two hundred thousand protein-caption pairs with rich functional notes for training. The finished model retrieves look-alike, like-meaning proteins from the main structure database. It also handled a harder test, searching across into an electron microscopy database it was not trained on, without any retraining. The authors argue such multimodal foundation models can speed up the interpretation of new structures from imaging methods like cryo-electron microscopy.

Catalog record

CREATORQifeng Wu, Zhengzhe Liu, Han Zhu, Yizhou Zhao, Daisuke Kihara, Min Xu, [3pt] Carnegie Mellon University, Purdue University
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE28 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-08023

Browse the full library · About the Tower of Babel