Grouping by MCS Example

Name: Grouping by MCS Example

Author: Ton van Daelen

Version: 1.0

Created: 8/2004

Modified: 6/2007 - updated all components to 6.1.1.

Purpose: Perform a maximal common subgraph search (MCSS) and use the results to cluster a compound library. This will ideally result in more chemically intuitive clustering.

Requirements: PP Server 6.1.1.

O/S: Any

Limitations: Note that the results can be quite sensitive to the setting of some input parameters. Experiment with these to get the best results. The MCSS is a slow step in this process. For a set of of 116 estrogen-antagonist it will take about 5 minutes.

Keyword: clustering, MCS, maximal maximum common substructure, subgraph, substructure, fragment, classpharmer

Contents: Grouping by MCS Example.xml

Installation:
NumberOfClusters (cluster component): number of groups to create for scaffolds.
MCSS Selection Criteria (main protocol): criteria for acceptings scaffolds returned by MCSS. The MCSS returns a large number of scaffolds that can be quite similar. The protocol then clusters these scaffolds, and selects for each cluster the scaffold with the largest number of rings that is found in the largest number of input samples. Other selection choices are possible and can be specified inside of the MCSS subprotocol. The protocol generates two sets of output: an html page with the clustered scaffolds and an html page with original library clustered by scaffold. It should be easy to extend the current protocol to use other sources for scaffolds.