Targetome Knowledge Graph for high-quality drug target interaction predictions

corresponding author: evansna@ohsu.edu

Setting up environment with conda/mamba:

(expectation of a cuda enabled GPU)

$ mamba env create -f environment.yml
$ mamba activate tkgdti 

# install tkgdti package
(kgdti) $ pip install -e . 

Downloading and processing datasets

We currently have methods to download the OGB-biokg dataset and the Hetero-A dataset.

(kgdti) $ python ./scripts/create_heteroa.py 
(kgdti) $ python ./scripts/create_biokg.py

Training a Complex^2 model

(tkgdti) $ python train_complex2.py --data /path/to/data/ --out /path/to/output 

Training a GNN model

(tkgdti) $ python train_complex2.py --data /path/to/data/ --out /path/to/output 

---

See /docs/ for additional details.