Paul Hawkins
OpenEye Scientific Software
When confronted with a problem in computational drug design there often exists a large number of tools that the computational chemist could apply to the problem. The difficulty lies in predicting which tool will be the most useful for that problem. There are a number of simple yet robust statistical tests that can easily be applied to help in deciding which tool is likely to perform better in prospective work, based on retrospective data. The application of a number of statistical tests to pose prediction and virtual screening data from the literature will be illustrated and the workshop participants will be provided with a set of methods to allow them to perform these tests on their own data. With an understanding of the utility of these techniques the workshop will then focus on virtual screening and lead-hopping and the use of appropriate statistical methods for these problems.
The application part of the session will focus on using shape-based methods for virtual screening and lead-hopping using the OpenEye tool ROCS. The workshop participant will learn about searching in shape space, using shape in virtual screening and the visual query editor vROCS. The vROCS editor will be used to generate queries for virtual screening or lead-hopping experiments. Using vROCS we will learn how to merge multiple molecules into a single query, edit molecules in a way that separates the structure of a molecule from the idea of a query, and validate the queries that class members generate in retrospective virtual screens. The workshop participants will learn about robust statistical methods that can be applied to virtual screening experiments and will use these methods to compare queries that they generate, enabling them to choose the best query for prospective experiments. Class members working in small groups will be able to apply shape-based virtual screening to their case studies.