takshan/mcanet

Source code of paper: "MCANet: Shared-weight-based MultiheadCrossAttention network for drug-target interaction prediction"

MCANet

Paper is available at Briefing in Bioinformatics

Dependencies

Quick install: pip install -r requirements.txt

Dependencies:

• python 3.8+
• pytorch >=1.2
• numpy
• sklearn
• tqdm
• prefetch_generator

Usage

python main.py <dataset> [-m,--model] [-s,--seed] [-f,--fold]

Parameters:

• dataset : DrugBank, Davis , KIBA , Enzyme , GPCRs or ion_channel
• -m or --model : select <model name> from MCANet, MCANet-B, onlyMCA or onlyPolyLoss, optional, default:MCANet
• -s or --seed : set random seed, optional
• -f or --fold : set K-Fold number, optional

Project Structure

• DataSets: Data used in paper.
• assets: Readme resources.
• utils: A series of tools.
• config.py: model config.
• LossFunction.py: Loss function used in paper.
• main.py: main file of project.
• model.py: Proposed model in paper.
• README.md: this file
• requirements.txt: dependencies file
• RunModel.py: Train, validation and test programs.

Citation

@article{10.1093/bib/bbad082,
    author = {Bian, Jilong and Zhang, Xi and Zhang, Xiying and Xu, Dali and Wang, Guohua},
    title = {MCANet: shared-weight-based MultiheadCrossAttention network for drug–target interaction prediction},
    journal = {Briefings in Bioinformatics},
    volume = {24},
    number = {2},
    pages = {bbad082},
    year = {2023},
    month = {03},
    issn = {1477-4054},
    doi = {10.1093/bib/bbad082}
}