This repository contains the code for the DDoS model implementation with PyTorch.
DDoS is a Graph Neural Network (GNN) based neural network model used for drug-drug-cell line synergy predictions.
Schwarz, Kyriakos, et al. “DDoS: A Graph Neural Network based Drug Synergy Prediction Algorithm.” arXiv preprint arXiv:2210.00802 (2022), https://arxiv.org/abs/2210.00802.
git clone
the repo andcd
into it.- Optionally, create a python
virtualenv
and install inside thepip
packages from therequirements.txt
file. Note thattorch
andCUDA
versions depend on the GPU/CUDA which you use locally.
Download Gene Expression
data (link: 1) and Drug Combination
data (link: 2). Place them into the data/preprocessing
directory.
- use
notebooks/DDoS_Preprocessing.ipynb
andnotebooks/DDoS_Dataset_Generation.ipynb
to generate DataTensors from the drug combination features (chemical structures and gene expression). - use
notebooks/DDoS-Hyperparam.ipynb
to find the best performing model hyperparameters. - use
notebooks/DDoS_Train_Eval.ipynb
to train / test on the best analyze the results.