This page organizes papers related to artificial intelligence published around the world. This page is summarized using Google Gemini and is operated on a non-profit basis. The copyright of the paper belongs to the author and the relevant institution. When sharing, simply cite the source.
Lightweight MSA Design Advances Protein Folding From Evolutionary Embeddings
Created by
Haebom
Author
Hanqun Cao, Xinyi Zhou, Zijun Gao, Chenyu Wang, Xin Gao, Zhi Zhang, Cesar de la Fuente-Nunez, Chunbin Gu, Ge Liu, Pheng-Ann Heng
Outline
PLAME is a lightweight MSA design framework that leverages evolutionary embeddings from pretrained protein language models for protein structure prediction to generate MSAs that perform better on low-homology and orphan proteins. PLAME balances agreement on conserved positions with the range of possible sequence variants through conservation-diversity loss. Beyond MSA generation, we developed an MSA selection strategy that filters high-quality candidates and a sequence quality metric that complements depth-based measures and predicts folding gains. On the AlphaFold2 low-homology/orphan benchmark, PLAME demonstrates state-of-the-art improvements in structural accuracy (e.g., lDDT/TM-score) and achieves consistent gains when used with AlphaFold3. PLAME also serves as a lightweight adapter that allows ESMFold to approach AlphaFold2 accuracy while maintaining comparable inference speed.
Takeaways, Limitations
•
Takeaways:
◦
Improving structural prediction accuracy for low-homology and orphan proteins by generating MSA using pretrained language models.
◦
Improving MSA quality through development of MSA selection strategies and sequence quality metrics.
◦
Demonstrated performance improvements in conjunction with AlphaFold2 and AlphaFold3.
◦
We demonstrate the potential for speeding up high-quality structural prediction using ESMFold.
