This repository is the implementation of the "PolyLLM: Polypharmacy Side Effect Prediction via LLM-Based SMILES Encodings" paper.

Start by cloning the repository, and installing the dependencies:

# Cloning
$ git clone https://github.com/sadrahkm/PolyLLM.git
 
# Installing dependencies
$ conda env create -f environment.yml
$ conda activate polyllm 

You then need to download the data, extract it and move it to the data directory):

$ wget https://snap.stanford.edu/biodata/datasets/10017/files/ChChSe-Decagon_polypharmacy.csv.gz
$ mkdir data/dataset
$ gunzip ChChSe-Decagon_polypharmacy.csv.gz
$ mv ChChSe-Decagon_polypharmacy.csv data/dataset 

You can now easily run the commands to prepare the data for inputting to the models:

$ python -m data.poly
$ python -m data.preprocessing

At this stage, you can run the MLP model with the generated data:

$ python -m mlp.mlp

If you would like to run the Graph Neural Network (GNN) model, you first need to generate embedding features and then feed them into the GNN:

# To generate embeddings
$ python embed.embedding

# To run the GNN model
$ python graph.graph