connorcoley / scscore

scscore

Description

The SCScore model assigns a synthetic complexity score between 1 and 5 to a molecule. The score is based on the premise that published reactions, overall, should exhibit an increase in synthetic complexity. The model has been trained on 12M reactions from Reaxys.

Usage

The standalone numpy model is defined in scscore/standalone_model_numpy.py

Dependencies if you want to use the final model

• RDKit (most versions should be fine)
• numpy

Dpendencies if you want to retrain on your own data

• RDKit (most versions should be fine)
• tensorflow (r0.12.0)
• h5py
• numpy