Skip to content

Latest commit

 

History

History
39 lines (32 loc) · 1.75 KB

README.md

File metadata and controls

39 lines (32 loc) · 1.75 KB

DrugOrchestra

DrugOrchestra is a multi-task learning neural model used for jointly training tasks of drug target prediction, drug response prediction, drug side effect prediction. It is proven to be better compared to single-task learning (training only one task) under the same training conditions. The paper is accepted by RECOMB2021. https://www.biorxiv.org/content/10.1101/2020.11.17.385757v1

Packages and environment

torch==1.4.0
torchvision==0.5.0
pandas==1.1.3
numpy==1.18.5
rdkit==2017.09.1
tqdm==4.48.0
scipy=1.4.1
sklearn==0.23.2
python==3.7.6
cuda==9.2

Data

The data after thresholding and feature extraction are available in
https://drive.google.com/file/d/1JvGDiNMAqWJb4Ya0c7ahV9fM1MCjtkxA/view?usp=sharing , with a README file describing each data file.

The data (numerical data) used for training and testing are available in
https://drive.google.com/file/d/1tzsZwk0exESwq1hoLii0SI5MWB-k5BxC/view?usp=sharingput .

How to run the code

Please refer to the https://github.com/jiangdada1221/DrugOrchestra/tree/master/MTL to see examples.

To make predictions

Run the script in 'Predict' folder by:

python make_prediction.py --SMILES arg1 --path_to_gnn_model arg2 --path_to_MTL arg3

arg1 is the input SMILES string you want to use to make predictions
arg2 is the path to the pretrained GNN model for drug embedding extraction
args is the path to the pretrained MTL model

By running this script, you will get output of prediced targets,response,side effects for the given drug

Correspondence

I check email frequently, if you have any question, feel free to contact [email protected]