Chapter 3. Model Building Techniques in Drug Discovery

Kimberly Crimin^[]

^[] Kimberly Crimin is Senior Principal Biostatistician II, Early Developement, Wyeth, USA.

Thomas Vidmar^[]

^[] Thomas Vidmar is Senior Director, Midwest Nonclinical Statistics, Pfizer, USA.

This chapter focuses on techniques used in early stages of drug discovery to optimize the model building process. The main topics discussed in this chapter are selecting training and test sets, selecting variables, and determining when the prediction of a new observation is valid. Statistical procedures that are used to build models are also reviewed. For each of the main topics we provide background, discuss a new procedure, and provide guidance on implementation in SAS. We apply these methods to a real drug discovery data set.