Understanding protein function with a multimodal retrieval-augmented foundation model

TL;DR

PoET-2 is a multimodal, retrieval-augmented protein foundation model that improves protein function prediction, especially for variants with multiple mutations, by integrating evolutionary constraints and structure information.

Contribution

This work introduces PoET-2, a novel multimodal, retrieval-augmented model with hierarchical transformers and dual decoders, advancing protein function understanding and variant effect prediction.

Findings

Achieves state-of-the-art zero-shot variant effect prediction.

Outperforms previous methods in small dataset supervised learning.

Excels at scoring complex mutations including indels.

Abstract

Protein language models (PLMs) learn probability distributions over natural protein sequences. By learning from hundreds of millions of natural protein sequences, protein understanding and design capabilities emerge. Recent works have shown that scaling these models improves structure prediction, but does not seem to improve mutation understanding and representation quality for protein function prediction. We introduce PoET-2, a multimodal, retrieval-augmented protein foundation model that incorporates in-context learning of family-specific evolutionary constraints with optional structure conditioning to learn generative distributions over protein sequences. PoET-2 uses a hierarchical transformer encoder that is equivariant to sequence context ordering and a dual decoder architecture with both causal and masked language modeling objectives, allowing PoET-2 to operate in both fully generative and bidirectional representation learning modes. PoET-2 achieves state-of-the-art performance on zero-shot variant effect prediction, excelling at scoring variants with multiple mutations and challenging indel mutations. In supervised settings, PoET-2 embeddings outperform previous methods for learning sequence-function relationships, especially with small datasets. This work highlights the benefits of combining retrieval augmentation with multimodal, family-centric modeling for advancing protein foundation models.