Strengthened GCN Model for Polymer Tg Prediction

This project contains graph con model which can be used to predict Tg of polymers. This project references the MPNN model from the open source project chemprop (https://github.com/chemprop/chemprop), as well as the CMC_GCN model from the paper Predicting Critical Micelle Concentrations for Surfactants Using Graph Convolutional Neural Networks（https://doi.org/10.1021/acs.jpcb.1c05264）.

Requirements

To use CMC_GCN model,you will need: python＞=3.7,CUDA＞=10.2

Install

Set up conda environment and clone the github repo

Create a new environment

$ conda create --name n python=3.7
$ conda activate n

Install requirements

$ pip install pytorch=1.9.1=py3.7_cuda10.2_cudnn7.6.5_0
$ pip install dgllife=0.2.6
$ pip install scikit-learn=1.0.2
$ conda install -c conda-forge rdkit=2020.09.1.0
$ conda install -c conda-forge tensorboard

Clone the source code of Strengthened GCN

git clone https://github.com/zeanli/Strengthened_GCN_Model_for_Polymer_Tg

Usage

To train the model, run

$ python workflow_test.py

To augment the data, run

python strengthen_ntimes.py

this will generate a csv file contained two columns To generate psmiles data, run

python strengthen_psmiles.py

To perform personalized differential augmentation, run

pythobn strengthen_personalized.py

this will generate a csv file contained 4 columns,the required dataset is the cmc_dataset_202.csv Three classic machine learning models are in the machine learning folder,you can change the data path and parameters as needed. To use MPNN model,You can follow the project documentation in https://github.com/chemprop/chemprop ,data needs to be replaced with the files in the dataset folder.