Today we'll be learning to how to use neural networks to predict a molecule's penetrance of the blood brain barrier (BBB) using smiles strings. You can read more about this here: https://pubs.acs.org/doi/10.1021/ci300124c. The goal of this recitation is to get some experience using sklearn / machine learning basics on some of the data we use in structural biology.
Make a copy of the notebook so your changes are saved!!
For Check:
- Load in the train and test data
- Use rdkit to get the fingerprints for your loaded molecules
- Pass the rdkit fingerprints into an MLPClassifier and create a loss curve
- In GradeScope, submit your plot and report the accuracy of your model
For Check+:
- Improve your test accuracy of the MLP by fiddling with model parameters like learning rate, batch size, etc.
- Upload your plots to GradeScope.