Tversky Similarity

Name: Tversky Similarity

Author: David Foss

Version: 2.0

Created: 1/2006

Modified: 9/2007 – Component upgraded, and help text revised/added.

Purpose: Conducts a Tversky similarity search for a set of target compounds against a set of reference compounds.

This is a specific implementation of the Molecular Similarity (Tanimoto, etc.) component using a UserDefined Coefficient to calculate the Tversky Similarity.

See Also:
Tversky Similarity Page at Daylight:
www.daylight.com/meetings/mug9...UG97/tv_tversky.html
Data Modeling Overview (Chapter 13 for Molecular Similarity documentation)

The Tversky Similarity component outputs the following properties:
* NumberOfClosest: Number of reference molecules within the specified range of minimum and maximum similarity.
* ClosestNames: Names of the closest (or nearest-neighboring) reference molecules within the specified range of minimum and maximum similarity.
* ClosestSimilarities: Similarity to the closest molecules.

* Pipeline Pilot and Daylight have different naming conventions for the different similarity contributions. Because of this, we have chosen different greek letters - Beta and Gamma -- for our Tversky parameters. Beta is the coefficient for SB and Gamma is the coefficient for SC.
* This is not a filter; all target compounds are passed through. However, some may not be within similarity range of any reference compounds and will have the property NumberOfClosest set to zero. You can filter on this, if necessary.
* Only fingerprint (or counts) properties are used in calculating the similarity. Properties of other types are ignored.


Requirements: Pipeline Pilot 6.1.1 or higher

O/S: PP Server Windows and Linux
PP Client Windows

Limitations: None

Keyword(s): Tversky, Molecular Similarity

Contents: Tversky Similarity.xml


Installation: Drag the component into the Pipeline Pilot client workspace and use in your protocols, or drag it and drop it on one of the Pipeline Pilot client explorer tabs to import it directly in the protocol database.