The official code implementation for DC_with_DDI_SupCon from the paper [Drug Combination prediction with Drug-Drug Interaction data as negative data and Supervised Contrastive Learning as pretraining technique (preprint)]
The overall framework of our work is shown below.
Step 1. Clone this repository and move to the directory.
git clone https://github.com/gujh14/DC_with_DDI_SupCon.git
Step 2. Install necessary packages.
pip install -r requirements.txt
Step 3. (If you want to use wandb for logging) Register your wandb profile on your local machine.
To preprocess and split data, run preprocess.ipynb and data_split_gcn.ipynb.
To train new model, run bash scripts (script_rw.sh and script_gnn.sh).
You need to adjust the arguments as you wish.
(If using wandb), use your own wandb entity & project name for arguments.
To visualize the embedding vectors of drug pairs, run visualization_embedding_space.ipynb.
To perform case studies, run case_study.ipynb.