Re: What do I do with the entries assigned to multiple categories? [message #81 is a reply to message #80] |
Sun, 26 July 2015 13:35 ![Go to previous message Go to previous message](/forum/theme/myThemeWithHFCSS/images/up.png) |
thomas
Messages: 693 Registered: June 2014
|
Senior Member |
|
|
I assume that you aim for visualizing the chemical space of your unique molecules and then use color to highlight, which regions of the space are covered by which source, possibly one plot per source. You certainly need to experiment a little with what works well for you data. What I can suggest is the following:
- make a copy of your file and (for simplicity) remove all columns except the chemical structure.
- remove rows with redundant structures and calculate the skeletonSpheres descriptor, which is the best for chemical similarity
- make a SOM (Data menu) with somewhat more reference vectors than compounds using the skeletonSpheres descriptor
- make a similarity analysis (Chemistry menu) with the skeletonSpheres descriptor
- make one or more scaffold analysis (with Murcko, most central ring system, ...)
- you may add labels to each scaffold with 'Add Row Numbers..." and choosing 'Use same number for same' Scaffold
- cluster dataset, if the number of compounds permit that
- save the file
Now you have a file with all compounds categorized in various aspects (scaffolds, cluster) and with x- and y-coordinates allowing to show the chemical space
- Now open your original file and select File->Merge, choose the previously created file, select the structure as merge key and add all other columns of that file to your original file.
- Create 2D-views to show the SOM space and Activity analysis space by using their x- and y-coordinates for the axes. You may quickly copy the views of the compound file with 'File->Open Special->Apply Template...' and choosing that file.
- Rows with the same structure appear in the views with the same coordinates, i.e. on top of each other. If you add a slight jittering, then you start to see multiple ones.
- To highlight all compounds of one category, e.g. all compounds from one source, you may use 'Set Focus...'. For that you either need to create a list for every category, or as a little trick, you set the focus to the selection. Then you create a new 2D-view with one axis unassigned and the other assigned to the category (e.g. source or scaffold number). In this view it is easy to select an entire category causing the selected rows to visually stick out in the other view that uses 'focus on selection'.
- You may use the 'category browser filter' and its dynamic feature to show the chemical space of the different sources one-by-one.
You could go a step further by taking the file with the unique compounds, make a copy, delete all rows except one of the scaffold columns, remove all redundant scaffolds and make a SOM or Activity analysis of that. This way you visualize scaffold space rather than compound space. If you save the file and merge it into your original using scaffold as key, you can do the same things with scaffold space that I have suggest above with compound space.
I hope you can derive something useful from these suggestions...
Thomas
|
|
|