De Novo Design Multiple Microplastic-Binding Peptides with a Protein Language Model-Guided Generative Adversarial Network

Overview

Folder	File Name	Description
data	PE.csv	Peptides affinity scores for PE plastic (obtained with PepBD)
data	PP.csv	Peptides affinity scores for PP plastic (obtained with PepBD)
data	PET.csv	Peptides affinity scores for PET plastic (obtained with PepBD)
data	collected_peptides.csv	Collection of the top 25% peptides from the three plastics
data	tok2idx.pkl	Mapping from amino acids to token IDs
code	utilities.py	Helper functions used across the project
code	models.py	AI models utilized in the project
code	train_ESM_predictor.py	Script to train the peptide affinity score predictor using ESM-2 embedding
code	train_PLMGAN.py	Script to train the PLM-GAN model proposed in the paper
results	GAN_peptides_top100.csv	Top 100 peptides generated by PLM-GAN (each peptide differs by at least 3 amino acids)
results	MD_simulation_results.xlsx	Equilibrium MD and Steered MD simulation results

Requirements

• torch
• fair-esm

Usage

1. Train the Instructors model with train_ESM_predictor.py
2. Train the PLM-GAN model with train_PLMGAN.py