Overview
The Structure Activity Landscape Index (SALI) is a numerical approach to quantifying activity cliffs [1] and is described in Guha and Van Drie (2008a). The paper describes an approach to visualize the activity cliffs in a SAR dataset in the form of a network (or graph) termed the SALI network. The original approach was based on static images generated using Graphviz. This visualization tool allows one to interactively explore a SALI network at different cutoff values, allowing one to focus on more or less significant activity cliffs. In addition the application now generates SALI curves (Guha and Van Drie, (2008b)), which allows one to characterize the ability of a modeling protocol to capture the SAR landscape.It is a modified version of the ZGRViewer application that visualizes networks described in the DOT language. The application has been tested on OS X, Linux and Windows Vista, though it should work anywhere where Java runs and Graphviz has been installed.
Update (06/02/09) - 1.9 uploaded which updates the version of zvtm. Properly handles SVG from recent versions of graphviz
Update (02/13/09) - 1.8.3 uploaded which properly renders aromatic rings
Update (02/12/09) - 1.8.2 uploaded with improved rendering and slightly faster fingerprints
Update (02/02/09) - 1.8.1 uploaded and the single jar file is much smaller than recent versions as it does not include the entire CDK
Update (01/26/09) - 1.8 uploaded and now brings back substructure highlighting. Also atoms are no longer colored, which makes it easier to see the highlighted substructure
Update (01/17/09) - 1.7 uploaded and includes improvements to fingerprint calculation performance and the latest 2D rendering code
Update (01/16/09) - Moved source code to GitHub repository, located at http://github.com/rajarshi/saliviewer/tree/master
Update (09/27/08) - Fixed a bug where the SALI curve UI was not synced with the chemical data that may have been freshly loaded. This update makes sure that when new chemical data is loaded, the SALI curve UI is killed.
Another result of this update is that the jar file is 15MB. This is because we now include the entire CDK library. This is done because the latest version does not properly display 2D depictions on OS X. After this is fixed we will provide a leaner update.
Update (05/13/08) - Updated to use dashed edges between nodes that have a similarity of 1.0
Update (05/08/08) - Updated so that the SALI zoom slider indicates the cutoff at which the network gets so large that layout and rendering might take a long time. Currently, a large network is defined as having 200 or more edges. The cutoff is rounded up to the nearest 10, so it's an approximate indicator
Update (05/06/08) - Fixed a bug which resulted in fingerprint calculation failing since the config file did not have the fingerprint type attribute in it. Also added a dialog to indicate that data loading is complete. Probably should make this into a message on the status bar
Update (04/26/08) - Fixed rendering issues with the latest CDK trunk. Also provided the option to choose which CDK fingerprinter the network should use
Update (04/25/08) - SALI curve functionality is now included in the application. The input format has also been extended to allow specification of one or more predicted property columns
Update (02/29/08) - Depictions of pairs of molecules are now resizable. The application also uses the graph directory specified in the Preferences window. Activities are formatted to two decimal places. Keyboard shortcuts added for loading SALI data and generating the SALI matrix. Also provide a menu option to toggle whether structure differences are highlighted
Update (02/17/08) - Depictions of individual molecules are now resizable
Update (02/16/08) - Structure depictions now display molecule name and activity in the panel itself, rather than just on the title bar. Updated preference pane for SALI operations. You can now alter the length and depth of the CDK fingerprint, highlight differences in the structure for a SALI pair (though for large molecules, MCSS detection may take a very long time) and also define whether a smaller activity value is better (say IC50) or a larger one is better (say -log IC50)
Update (02/15/08) - Updated code to handle molecule names starting with a digit or names containing dashes. The input file format has been changed and the code updated to support it. No longer crashes if activity is missing. Edges corresponding to cliffs for which the similarity equals 1.0, have diamond shaped arrow heads. All other edge will have the normal triangular arrow head.
Update (01/23/08) - Structural differences between pairs are not highlighted due to a bug in the CDK MCSS detection procedure. Once that's fixed, it'll be back
Update (01/22/08) - The program will now highlight the difference in the two structures when an edge is clicked. Note that this seems to be buggy as the MCSS algorithm is sometimes behaving weirdly. Also, edges of the graph are now colored based on the magnitude of the activity cliff, at a given cutoff value. Black is the largest cliff and light grey is the smallest cliff.