A Comprehensive Comparative Analysis of Deep Learning Based Feature Representations for Molecular Taste Prediction

Papar | Citation

This repository contains the official code for this paper:A Comprehensive Comparative Analysis of Deep Learning Based Feature Representations for Molecular Taste Prediction

Data

The 'tastes_split' folder contains three types of dataset used in our experiment.

Performance

we tested 14 models,including six fingerprint models, three CNN-RNN models and five GNN models in our dataset. The following table and figure are one of the three types of tastes.The model implementation is supported by DeepPurpose package.

The 'result' folder contains the data from our experiment along with instructions on how to calculate metrics and create bar graphs.

Showcases of AUROC scores

Citation

If you find this repository useful, please cite our paper using

@Article{foods12183386,
    AUTHOR = {Song, Yu and Chang, Sihao and Tian, Jing and Pan, Weihua and Feng, Lu and Ji, Hongchao},
    TITLE = {A Comprehensive Comparative Analysis of Deep Learning Based Feature Representations for Molecular Taste Prediction},
    JOURNAL = {Foods},
    VOLUME = {12},
    YEAR = {2023}
}