Dataset information

Overview

This project uses four benchmark datasets for drug-target interaction (DTI) prediction:

• BindingDB
• BioSNAP
• Human
• DrugBank

All datasets were obtained from their official sources or official benchmark release pages and were further processed into a unified format for model training and evaluation.

Data Sources

1. BindingDB

BindingDB is a public database of experimentally measured binding affinities between proteins and small molecules.

• Official website: https://www.bindingdb.org/
• Official download page: https://www.bindingdb.org/rwd/bind/chemsearch/marvin/Download.jsp

2. BioSNAP

BioSNAP provides curated biomedical network datasets. In this project, we use the drug-target interaction benchmark released through the Stanford SNAP BioSNAP collection.

• Official BioSNAP dataset page: https://snap.stanford.edu/biodata/datasets/10015/10015-ChG-TargetDecagon.html
• BioSNAP collection page: https://snap.stanford.edu/biodata/

3. Human

The Human dataset used in this project is a widely adopted DTI benchmark dataset from the benchmark release used in prior DTI studies.

• Official benchmark release source: https://github.com/lifanchen-simm/transformerCPI
• Human dataset file: https://github.com/lifanchen-simm/transformerCPI/blob/master/Human%2CC.elegans/dataset/human_data.txt

4. DrugBank

DrugBank is a comprehensive knowledgebase integrating drug information with drug-target associations.

• Official website: https://go.drugbank.com/
• Official release/download page: https://go.drugbank.com/releases/latest

Download Policy

The raw datasets were obtained from the official websites or official benchmark release pages listed above.

Directory Structure

Dataset/
├── bindingdb/
├── biosnap/
├── human/
├── drugbank/
└── README.md