Whenever you have a dataset of known actives for a given target and a reservoir of potential decoys, you can apply the MUV workflow to design Maximum Unbiased Validation datasets for your specific VS validation application.

The workflow is based on the MATLAB Spatial Statistics Toolbox, available from this site. (Download) As a consequence, you need to have a valid license of The Mathworks MATLAB for the workflow to function. A re-implementation based on open source software is currently under way. The toolbox is complemented by a JAVA library jMUV.jar managing SD-file I/O and descriptor calculation. jMUV itself is based on the Chemistry Development Kit CDK. (Version 1.0.3, Download) Both libraries have to be included in your MATLAB classpath.

The tutorial HowTo-MUV provides a step-by-step guide for installing the necessary software and generating MUV datasets. Two example datasets in SDF-format used in the tutorial are also available for download below.

Please also have a look at the MUV FAQ.

Downloads:

File	Version, Date	Comment
SpSt.tar.gz	0.9, 1 Oct 2008	Spatial Statistics Toolbox for MATLAB
jMUV.jar	0.9, 1 Oct 2008	JAVA Helper Library for the Spatial Statistics Toolbox
ckd-1.0.3.jar	1.0.3, 5 May 2008	Chemistry Development Kit (Open Source Chemistry Toolkit for JAVA)
examples.tar.gz		Example datasets for MUV benchmark set generation (80MB)