3.4. Modeler Theory and Details

The sections in this chapter are intended to provide users with deeper and more detailed insights into the modeling methods we use and the science behind them. The topics are arranged in an order that roughly parallels the usual ADMET Modeler workflow:

Modeling Methods contains detailed descriptions of each available type of modeling method available.
Ensembles describes how individual models are selected for inclusion in an ensemble and how ensemble outputs are generated.
Descriptor Selection discusses how the input descriptors are sorted for incremental inclusion in models.
Kohonen Mapping provides a detailed description of how our preferred training and test set selection method works.
Confidence Analysis presents the theoretical underpinnings of the method we use to estimate how confident one should be in individual ANNE classifications.
Finding the Best Ensemble describes how multiple performance metrics are combined when applying our automatic model selection tool.
DELTA Models explains the logic behind constructing DELTA Models, which is the recommended way to augment our global models with your in-house data.

Available descriptors are enumerated in 2.10.4: Descriptor Definitions, which includes brief descriptions of the quantities they represent.