TarIKGC: a target identification tool based on semantics enhanced knowledge graph completion

• This repository contains the source code ,the data and trained models.

Train

Model training can be started by running the main.py script:

python main.py  --gpu 2  --epoch 40

Notes:

Pre-download and Pre-generate Necessary Files to Save Model Training Time:

1. Download the disease embedding file from this link. Move the downloaded file to the dataset/ directory and then execute the script ICD_embedding.py.

2. Run the script preprocess.py to generate the mol_feature.pt file.

Training the model will display folder output/ with the following structure:

KG
└── |"{now_time}_max{self.p.max_epochs}
        ├── models
        ├── results

• You can run the reposition_lab.py script to predict the targets of a molecule of interest. you will obtain a target recommendation list. For example, You can reproduce the prediction results of the CDK2 inhibitor discovery by running the following scripts:

python reposition_lab.py --gpu 0