FOM: Figure of Merit
Parameter Information
Sample Selection
The sample selection option indicates whether to cluster genes or experiments.
FOM Iteration Selection
The FOM module can be run several times to get average FOM values. This feature
is useful for KMC FOM runs since each KMC run gives possibly unique results
due to the random nature of KMC intialization. If this parameter is greater
than one the FOM graph will reflect the mean FOM values with SD bars.
Algorithm Selection Tabs
The Figure of Merit is, in concept, a measure of fit of
the expression patterns for the clusters produced by a particular
algorithm. MeV's FOM implementation provides FOM results for
running the KMC and CAST clustering algorithms. Each algorithm
is initialized by selecting either the K-Means/K-Medians tab
or the CAST tab.
KMC Parameters
Means/Medians Option
The Means or Medians option indicates whether each cluster's centroid vector
should be calculated as a mean or as a median of the member expression patterns.
Maximum Number of Clusters
This positive integer value indicates the maximum number of clusters to be created.
For instance, if the entered value is 10 then KMC is run 10 times to produce 1,2,3..,10
clusters. An FOM value is returned for each run.
Maximum Number of Iterations
This positive integer value is the maximum number of times that all the elements in the data set
will be tested for cluster fit within a single KMC run. On each iteration each element
is associated with the cluster with the closest mean (or median).
Note that a KMC run will terminate when either no elements
require migration (reassignment) to new clusters or when the maximum number of
iterations has been reached.
CAST Parameters
Threshold Interval
*For FOM an interval is used to perform a series of CAST runs in which the Affinity Threshold
is incremented from 0.0 by the interval indicated. The default of 0.1 is often a good value
since it provides 11 CAST results from 0.0 to 1.0 incremented by 0.1.
The threshold parameter is a value ranging from 0.0 to 1.0 which is used as a
cluster affinity threshold. Each expression element will have an affinity for the current cluster being
created based on it's relationship to the elements currently in the cluster. If that affinity
is greater than the supplied threshold the gene is permitted to be a member of the cluster.