PLNet: Persistent Laplacian Neural Network for Protein-Protein Binding Free Energy Prediction

Description

This manual is for the code implementation of paper "PLNet: Persistent Laplacian Neural Network for Protein-Protein Binding Free Energy Prediction"

Comments

• Primary dataset: place P2P.csv in the data/ directory (i.e. data/P2P.csv).
• The repository includes tooling and an optional virtual environment under env/; add generated environments and other local artifacts to .gitignore so they are not committed.
• Start by running src/config.py — setup and configuration instructions will be printed to the terminal.
• The bin/ directory holds external executables (e.g. JACKAL, SCAP, ProFix). You can populate bin/ yourself following the instructions printed by src/config.py, or request prebuilt binaries from xingjianxu@ufl.edu.
• All datasets support feature extraction including ESM embeddings, biophysical properties, and pairwise similarity matrices

Dependencies

• fair-esm (2.0.0): For protein language modeling
• torch (2.0.0): For deep learning operations
• biopython (1.79): For sequence analysis
• numpy (1.21.0): For numerical computations
• pandas (1.3.0): For data manipulation
• requests (2.26.0): For PDB file downloading
• tqdm (4.62.0): For progress tracking
• ProFix: For PDB file processing
• Jackal: For protein structure analysis
• SCAP: Side Chain Analysis Program for protein mutation generation (must be installed in system PATH)

Usage

Install Environment and package

python src/config.py

Then redo again for install package.

Data Generation for P2P

python src/data_preparation.py  # Prepare all P2P and generate the feature for ESM and Persistent Homology