This project includes the training process of multi-modal polymer property prediction.
- python
- pytroch
- RDKit
- numpy
- pandas
Tips: Using code conda install -c rdkit rdkit
can help you install package RDKit quickly.
The install time is near 10-15 minutes.
For multi-modal polymer representations:
1.Descriptors
We use Mordred to calculate various of monomer descriptors (https://github.com/mordred-descriptor/mordred). Then we apply a feature downselection process to filter the optimized descriptors (https://github.com/PV-Lab/MLforCOE).
2.Sequence and Graph Representations
The construction of sequence and graph representations can be found in ./chemreprop/features/seq_featurization.py
and ./chemreprop/features/featurization.py
Demo training and predicting codes can be seen in ./run.sh
We thank all cited codes in github, which are all quite helpful for realizing our work.