PIGLET

Proteome-wide Interaction Graph Link Prediction by Embedding with Transformers

Kristy A. Carpenter and Russ B. Altman

This repository contains the code required to run the PIGLET method introduced in the manuscript "Drug-Target Interaction Prediction with PIGLET."

Dependencies

• networkx==3.4.2
• numpy==1.26.4
• python==3.10.19
• scipy==1.15.3
• torch==1.13.1+cu117
• torch-cluster==1.6.3
• torch-geometric==2.7.0
• torch-scatter==2.1.2
• torch-sparse==0.6.18
• torch-spline-conv==1.2.2+pt113cu117
• torchmetrics==1.8.2
• tqdm==4.67.1
• wandb==0.23.1

We recommend running this code with CUDA version 11.7 and GCC version 10.1.

Setup

Please download the accompanying data from Zenodo here and move everything to the data directory. mean_canonical_embs.tar.gz should be unpacked to a directory: data/mean_canonical_embs

Usage

Navigate to the scripts directory and run main.py with the desired arguments to train a PIGLET model, e.g. python main.py --split drug --drugbank_mp --cv_fold 0