Related Experiment Video
Updated: Jul 9, 2025
09:37
An Integrated Approach for Microprotein Identification and Sequence Analysis
Published on: July 12, 2022
3.4K
Semantic search using protein large language models detects class II microcins in bacterial genomes
Anastasiya V Kulikova1, Jennifer K Parker2, Bryan W Davies2,3
1Department of Integrative Biology, University of Texas at Austin, Austin, Texas, USA.
Biorxiv : the Preprint Server for Biology
|November 28, 2023
Summary
Large language models identify novel antimicrobial peptides (microcins) missed by traditional methods. This AI approach enhances the discovery of potential new antibiotics from bacterial genomes.
Area of Science:
- Microbiology
- Bioinformatics
- Computational Biology
Background:
- Class II microcins are antimicrobial peptides with potential as novel antibiotics.
- Discovery of new microcins is challenging due to their short length and sequence divergence.
- Existing methods like BLAST struggle to identify all relevant microcins.
Approach:
- Utilized numerical embeddings from protein large language models (LLMs) to detect microcins.
- Applied LLM embeddings to bacterial genome assemblies.
- Compared the performance of LLM embeddings against sequence-based methods like BLAST.
Key Points:
- LLM embeddings significantly outperform BLAST in reliably detecting known class II microcins.
- Microcins with high sequence divergence cluster closely in the LLM embedding space.
- Novel putative microcins were identified in *Escherichia coli*, *Klebsiella* spp., and *Enterobacter* spp. genomes.
Conclusions:
- Protein LLM embeddings offer a powerful new tool for discovering antimicrobial peptides.
- This AI-driven approach overcomes limitations of sequence-based methods for microcin identification.
- The findings expand the known repertoire of class II microcins and antibiotic discovery potential.
