3 Ligand-based Screening: Machine Learning
This chapter covers
- The end-to-end process of a machine learning project in the context of cardiotoxicity prediction.
- How to acquire, curate, and standardize molecule datasets.
- Training and evaluating a linear model, which we can save for later use.
- How to improve our model with regularization and non-linear transformations.
- Hyperparameter tuning with grid search and randomized search.
Last chapter, we learned about compound filters and similarity searching in the context of ligand-based virtual screening. In this chapter, we will review one way that ML fits into our virtual screening pipeline. The key stages in our workflow are illustrated in figure 3.1: