QSAR
Python2.7, Scikit-learn, RDkit
Descriptors: 196 RDkit descriptors
- First, a set of machine learning methods trained by a curated set of BACE inhibitor retrieved from CHEMBLE database. The ligands with either ambiguous activity or reported activity for a condition rather than a BACE single protein were omitted from the data set. IC50 activities were converted to –log IC50 prior to model development.
- Descriptors: 196 RDkit descriptors were generated
- Pre processing: correlated variables were removed.
- Models: A MLP neural network were used for the model building.
- Then the model were used to predict the activity the 154 ligands.
No