Workflow

Workflow:
Split dataset into train/test/dev --> [Done in data_preprocessors.py]
Convert all BigSMILES strings to SMILES --> [Done in  data_preprocessors.py]
Convert all SMILES strings into graphs using existing code --> [Started in  data_preprocessors.py]
  - Add features to all BigSMILES nodes indicating they are blocks --> [Done]
  - Convert graphs to DeepSNAP graphs --> [Done]
  - Add Mn, T, f_A features to graph --> [Done]
Run GAT/GCN for classification

Blog Link:
https://medium.com/@littlethomas98/block-copolymer-phase-prediction-using-graph-neural-network-architectures-3ebec4f6c310



Goal: 
recreate phase diagram for copolymers using GNN

Database: 
https://github.com/olsenlabmit/BCDB
Holds 5300 block copolymer melt phase measurements. Copolymer structure is stored in BigSMILES

use https://www.rdkit.org/docs/source/rdkit.Chem.rdmolfiles.html (specifically .MolFromSmiles to convert from SMILES to usable input maybe?)
look at https://github.com/coleygroup/polymer-chemprop-data/blob/main/datasets/vipea/make-polymer-fps.py for inspiration

also look at https://github.com/olsenlabmit/BigSMILEStoTuples for converting BigSmiles to tuples