The rise of antibiotic-resistant bacteria drives an urgent need for novel antimicrobial agents. Antimicrobial peptides (AMPs) show promise due to their multiple mechanisms of action and reduced propensity for resistance development. Here we present LLAMP (Large Language model for AMP activity prediction), a target species-aware AI model that leverages pre-trained language models to predict minimum inhibitory concentration (MIC) values of AMPs. Through screening approximately 5.5 million peptide sequences, we identified peptides 13 and 16 as the most selective and most potent candidates, respectively. Analysis of attention values allowed us to pinpoint critical amino acid residues (e.g., Trp, Lys, and Phe). Using this information, we enhanced the amphipathicity of peptide 13 through targeted modifications, yielding peptide 13-5 with the highest antimicrobial activity among the variants. Notably, peptides 13-5 and 16 demonstrated antimicrobial potency and selectivity comparable to the clinically investigated AMP pexiganan. Our work demonstrates AI's potential to accelerate peptide-based antibiotic discovery.
You can use peptide tuned ESM-2 model weight here.
https://huggingface.co/Daehun/peptide_tuned_ESM-2
You can download LLAMP model weight here.
https://drive.google.com/file/d/1hmfL7uRZsHo4pn0o0nqaqPcntxGJIhE7/view?usp=sharing
The model implemented for the AMP MIC prediction in inference.ipynb.
# environment setting
$ git clone https://github.com/GIST-CSBL/LLAMP.git
$ cd LLAMP
$ conda create -n LLAMP python==3.9.13
$ conda activate LLAMP
$ pip install -r requirement.txt
$ pip install ipykernel
$ python -m ipykernel install --user --name LLAMP --display-name "LLAMP"
Daehun Bae ([email protected])
Hojung Nam* ([email protected])
*Corresponding Author