<div align=center><img width="800" height="450" src="https://github.com/ShengtianSang/SemaTyP/blob/main/figures/Illustration_of_SemKG.jpg"/></div>

SemaTyP: a knowledge graph based literature mining method for drug discovery

This is the source code and data for the task of drug discovery as described in our paper: "SemaTyP: a knowledge graph based literature mining method for drug discovery"

Requirements

• scikit-learn
• numpy
• tqdm

Data

In order to use the code, you have to provide

• Theraputic Target Database You don't need to download by yourself, I have uploaded all the TTD 2016 version in <./data/TTD>.
• SemedDB You need to download from here with password:1234 to obtain the whole knowledge graph. After downloading the "predications.txt" file, please replace the file <./data/SemedDB/predications.txt>. with this new downloaded file.

Run the codes

Install the environment.

pip install -r requirements.txt

Construct training and test data.

python experimental_data.py

Train and test the model.

python main.py

Illustration of feature selection

<div align=center><img width="800" height="300" src="https://github.com/ShengtianSang/SemaTyP/blob/main/figures/Illustration_of_Feature_selection.jpg"/></div> <p align="center"> An illustration of the features constructed in our work. </p>

File declaration

data/SemmedDB： contains all relations extracted from SemmedDB, which are used for constructing the Knowledge Graph in our experiment. The whole "predications.txt" contains 39,133,975 relations, we just leave a small sample "predications.txt" file here which contain 100 relation. The whole "predications.txt" file coule be downloaded from

data/TTD： contains the drug, target and disease relations retrieved from Theraputic Target Database.

experimental_data.py: constuct the drug-target-disease associations from TTD and Knowledge Graph.

knowledge_graph.py: construct the Knowledge Graph used in our experiment.

data_loader.py：used to load traing and test data.

main.py：used to train and test the models

Cite

Please cite our paper if you use this code in your own work:

@article{sang2018sematyp,
  title={SemaTyP: a knowledge graph based literature mining method for drug discovery},
  author={Sang, Shengtian and Yang, Zhihao and Wang, Lei and Liu, Xiaoxia and Lin, Hongfei and Wang, Jian},
  journal={BMC bioinformatics},
  volume={19},
  number={1},
  pages={1--11},
  year={2018},
  publisher={Springer}
}