1zzt / HiGPPIM

HiGPPIM - A Hierarchical Graph Neural Network Framework for Predicting Protein-Protein Interaction Modulators with Functional Group Information and Hypergraph Structure

1. overview

2. Prerequisites

• python 3.9.0
• pytorch 1.12.0
• networkx 3.1
• numpy 1.23.5
• pandas 2.0.0
• rdkit 2022.9.5
• torch-cluster 1.6.0+pt112cu113
• torch-geometric 2.3.0
• torch-scatter 2.1.0+pt112cu113
• torch-sparse 0.6.16+pt112cu113

3. Datasets

For PPIM identification tasks, PPI families include Bcl2-Like/Bak-Bax, Bromo Domain/Histone, CD4/gp120, LEDGF/IN, LFA/ICAM, Mdm2-Like/P53, Ras_SOS1, and XIAP/Smac.

For PPIM potency prediction tasks, PPI families include Bcl2-Like/Bak-Bax, Bromo Domain/Histone, CD4/gp120, LEDGF/IN, LFA/ICAM, Mdm2-Like/P53, and XIAP/Smac.

4. Usage

python main.py --dataset bcl2_bak --task classification --batch_size 32 --lr 0.0001 --gpu 0