This repository contains a suite of Large Language Models (LLMs) designed for high-throughput mining and generation of antimicrobial peptides (AMPs). Each model serves a unique purpose in the discovery and evaluation of potential AMPs:

• ProteoGPT: A pre-trained model for generating and analyzing amino acid sequences.
• AMPGenix: A sequence generator capable of producing potential AMP sequences.
• AMPSorter: A classifier designed to identify AMPs from peptide datasets.
• BioToxiPept: A classifier aimed at determining the cytotoxicity of short peptides.

All models and config files can be downloaded from here.

Clone this repository to your local machine using:

git clone https://github.com/W1V1995/AMP_Project.git

Download the AMP_models and place it in the AMP_Project directory.

Navigate into the cloned directory:

cd AMP_Project

Ensure that all dependencies are installed by following the installation instructions provided in the requirements.txt file or the dedicated installation guide.

To create a classifier by fine-tuning, execute the following command:

sh Fine-tuning_classifier.sh

To create a generator by fine-tuning, execute the following command:

sh Fine-tuning_generator.sh

Parameters such as batch_size, epochs, etc., and output path can be customized.

To generate sequences using AMPGenix, run:

sh AMPGenix.sh

Parameters such as ntokens, nsamples, prefix, model_path, save_samples_path, etc., can be adjusted as per your requirements.

For identifying AMPs from a peptide dataset using AMPSorter, execute:

sh AMPSorter_predictor.sh

Customize parameters including batch_size, raw_data_path, model_path, classifier_path, output_path, candidate_amp_path, etc., to fit your dataset and path.

To predict the cytotoxicity of short peptides with BioToxiPept, use:

sh BioToxiPept.sh

Adjustable parameters are batch_size, raw_data_path, model_path, classifier_path, output_path, candidate_pep_path, etc.

To predict the antimicrobial activity of short peptides based on charged residues and hydrophobic residues, use:

python QSAR.py

Adjustable parameters are samples_path, output_path.

To utilize AmpSorter or BioToxiPept for predictions, prepare a CSV file containing your sequence data. Ensure the file includes a column named "Sequence". Example format:

Sequence
<sequence_1>
<sequence_2>
...

Save your dataset in the format /Data/Sequence.csv or modify the script parameters to point to your custom data path.