niharikavadlamudi / d4_multi_task_learning_molecular_property_prediction

Baseline Experiments for Multi-Task Molecular Property Prediction

Multi Task Learning for Molecular Property Prediction

Description : This repository will consist baseline experiments for Multi-Task Molecular Property Prediction on Harvard's TDC dataset .

Installation

Easy installation via conda :

conda env create --file d4_mtp.yml --python=3.9
conda activate d4_mtp

TDC Regression Tasks

A total of 6 tasks - Caco-2 , Lipophilicity , Solubility (AqSolDB) , PPBR , Acute Toxicity LD50 & Clearance (Hepatocyt) are catergorised under regression. To generate the datafiles ( train & test ) , run the following commands :

cd tdc_regression
bash reg_data_generation.sh 

For training the network :

cd tdc_regression
bash train_reg_model.sh 

TDC Classification Tasks

A total of 4 tasks - Bioavailability,CYP P450 2D6 Inhibition,Ames Mutagenicity & hERG Blockers are catergorised under classification. To generate the datafiles ( train & test ) , run the following commands :

cd tdc_classification
bash clf_data_generation.sh 

For training the network :

cd tdc_classification
bash train_clf_model.sh 

Weights & Biases Page

A one-stop portal for observing current experiments :

• Regression MTP Experiments
• Classification MTP Experiments

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change. Please make sure to update tests as appropriate.

License

MIT