A PLMs based protein retrieval framework

2024-07-16

arXiv.org

AI Summary

Existing protein retrieval methods (e.g., BLAST) rely heavily on sequence similarity, limiting their ability to identify evolutionarily distant proteins with conserved structure or function. This work introduces the first dense retrieval framework for proteins based on protein language models (PLMs), leveraging ESM and ProtBERT to generate semantically informed embeddings that explicitly decouple sequence similarity bias. The approach integrates these embeddings with efficient approximate nearest neighbor indexing (e.g., FAISS) to enable scalable vector-based retrieval. Its core contribution is the first application of PLM-derived embeddings for functional, cross-sequence protein retrieval—enabling semantic matching at both structural and functional levels. Evaluated on multiple benchmarks, the method significantly improves recall of remote homologs and functionally similar proteins, successfully recovering critical functional proteins missed by BLAST. It establishes a new paradigm for protein functional annotation and novel target discovery.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Image and Video Retrieval
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Protein retrieval, which targets the deconstruction of the relationship between sequences, structures and functions, empowers the advancing of biology. Basic Local Alignment Search Tool (BLAST), a sequence-similarity-based algorithm, has proved the efficiency of this field. Despite the existing tools for protein retrieval, they prioritize sequence similarity and probably overlook proteins that are dissimilar but share homology or functionality. In order to tackle this problem, we propose a novel protein retrieval framework that mitigates the bias towards sequence similarity. Our framework initiatively harnesses protein language models (PLMs) to embed protein sequences within a high-dimensional feature space, thereby enhancing the representation capacity for subsequent analysis. Subsequently, an accelerated indexed vector database is constructed to facilitate expedited access and retrieval of dense vectors. Extensive experiments demonstrate that our framework can equally retrieve both similar and dissimilar proteins. Moreover, this approach enables the identification of proteins that conventional methods fail to uncover. This framework will effectively assist in protein mining and empower the development of biology.

Problem

• Protein Retrieval
• Sequence Similarity
• Functional Similarity

Innovation

• Protein Language Model
• Digital Representation
• Smart Database