The SOM wizard steps you through the process of creating Self-Organizing Maps.

SOMs are generated either through the creation of Quicklists that contain objects with numeric or structural data or by isolating a set of results data. All objects have to be the same basic type, such as compounds or batches. If the quicklist contains multiple types of dataobjects, you will need to select which type of data that you would like to use to create the map.

You can make a SOM in several ways:
QuickList Method

Create a new quicklist by clicking on the '+' button in the lower left hand corner of SLIMS and specify name for the new quick list such as 'SOM Data for Indoprofen' and then filling the quicklist with compounds or batches.
Goto the compound or batch table screen and drag the desired compounds to the newly created quicklist. The quicklist will blink indicating that the compounds have been successfully placed. You can also navigate to a results page and select batch results. This is a good way of organizing collections of hit compounds.
Goto to the SOM inventory objects and click on the "Make SOM" button in the command pane. This will begin the SOM creation wizard that will walk you through the process. Note: You will need to select the quicklist that you just created from the second wizard page.

Results Method

Go to a results set and select a tab that contains batches. You can then isolate an interesting set of compounds either by filtering the data to only include compound above or below a certain threshold.
Click on Analyze and select Create Self-Organizing Map.

To see the exact steps to create a som, please see the SOM section of the Spinal Muscular Atrophy dataset tutorial.

SOM Creation Options

The options for creating a SOM are given on the following wizard page:

Option	Description
Name your SOM	Enter a descriptive name of the Self Organizing Map
iterations	Generating a SOM requires several iterations of placing the target compounds on the map and allowing the map's nodes to adjust itself and it's neighbors to become even more similar to the compounds that the nodes are most similar. The number of iterations controls the number of times that this happens for all compounds.
Learning Rate	The learning rate controls how quickly each node becomes more similar to compounds that it is closest to during each iteration. A value of 1 means that each time a node is the most similar to a given compound, then it will adjust itself to be exactly like the compound. In general, slower learning rates allow more compounds to be attracted to a given node. Although, slower learning rates usually require more iterations.
Learning Radius	In addition to a node becoming more similar to a compound, it can drag it's neighbors along for the ride. In this fashion, not only do structurally similar compounds become attracted to a node, but the surrounding nodes also contain groups of compounds similar to their neighbors. A larger learning radius will dramatically slow down the SOM generation.
Quick SOM	Using the quick som option restricts compounds to only be checked in neighborhoods that they were previously attracted to. In most cases, the quick SOM will have very little effect, but in some cases it might not group structures correctly.
Randomize	The order of compounds used in the training process can generate different Self-Organizing Maps. It is usually a good idea to randomize the presentation.
Maximum Compounds	Sometimes you might select a large number of compounds to build a Self-Organizing Map from. You might want to select a number of training compounds to build your map as opposed to generating the map on all compounds. You will be able to place the remaining compounds on the map at a later time.
Map size rows	The map size controls the number of nodes in the Self-Organizing map. A larger map takes longer to generate and will produce smaller clusters (in general). Sometimes it is useful to generate a smaller map and then progressively make the map size larger.
Map size cols	See map size rows
Help	Displays this page.

This HTML Help has been published using the chm2web software.