2 |
26 Feb 07 |
jari |
1 |
<html> |
2 |
26 Feb 07 |
jari |
2 |
<body bgcolor = "#FFFFCC"><basefont face = "Arial"> |
2 |
26 Feb 07 |
jari |
3 |
<h1>Centroid Variance or Entropy Cluster Selection</h1> |
2 |
26 Feb 07 |
jari |
4 |
<hr size = 10> |
2 |
26 Feb 07 |
jari |
5 |
|
2 |
26 Feb 07 |
jari |
6 |
The cluster centroid is the mean expression vector of a cluster. Often the centroid is used |
2 |
26 Feb 07 |
jari |
7 |
to characterize changes in gene expression for a set of elements in a cluster. |
2 |
26 Feb 07 |
jari |
8 |
Centroid Entropy or Variance Ranking cluster selection is an algorithm used to rank a |
2 |
26 Feb 07 |
jari |
9 |
set of clusters based on Centroid variability and then select candidate clusters meeting |
2 |
26 Feb 07 |
jari |
10 |
supplied criteria. |
2 |
26 Feb 07 |
jari |
11 |
This process will tend to find clusters that meet a minimum size and have relatively |
2 |
26 Feb 07 |
jari |
12 |
higher centroid variability over the expression measurements. The selected clusters will tend |
2 |
26 Feb 07 |
jari |
13 |
to have centroids that vary greatly over the measurements. Be certain to understand how |
2 |
26 Feb 07 |
jari |
14 |
using <i>Centroid Variance</i> versus using <i>Centroid Entropy</i> will affect the outcome. |
2 |
26 Feb 07 |
jari |
15 |
|
2 |
26 Feb 07 |
jari |
16 |
<h1>Parameters</h1> |
2 |
26 Feb 07 |
jari |
17 |
<h2>Desired Number of Clusters</h2> |
2 |
26 Feb 07 |
jari |
18 |
This parameter indicates the number of clusters that should be selected from the input set. |
2 |
26 Feb 07 |
jari |
19 |
If during execution it turns out that the number of input clusters is smaller than the |
2 |
26 Feb 07 |
jari |
20 |
number of clusters desired, then all input clusters are returned as the result. |
2 |
26 Feb 07 |
jari |
21 |
|
2 |
26 Feb 07 |
jari |
22 |
<h2>Minimum Cluster Population (# of elements)</h2> |
2 |
26 Feb 07 |
jari |
23 |
The minimum cluster size describes the minimum number of genes or experiments that should |
2 |
26 Feb 07 |
jari |
24 |
be in the cluster. In some cases clusters may have low variability but are only composed of |
2 |
26 Feb 07 |
jari |
25 |
a couple of elements. |
2 |
26 Feb 07 |
jari |
26 |
|
2 |
26 Feb 07 |
jari |
27 |
<h2>Rank Clusters on Centroid Variance</h2> |
2 |
26 Feb 07 |
jari |
28 |
This measure takes the sum of squared errors for the centroid vector where the elements are the |
2 |
26 Feb 07 |
jari |
29 |
centroid values and each is compared to the centroid's mean value. |
2 |
26 Feb 07 |
jari |
30 |
|
2 |
26 Feb 07 |
jari |
31 |
<h2>Rank Clusters on Centroid Entropy</h2> |
2 |
26 Feb 07 |
jari |
32 |
Centroid entropy is a measure of the dispersion of centroid values over the observed range of |
2 |
26 Feb 07 |
jari |
33 |
centroid values. In this case, the high entropy centroids would have values that are widely dispersed |
2 |
26 Feb 07 |
jari |
34 |
between the extremes of the centroid. Note that the range of centroid values could be rather |
2 |
26 Feb 07 |
jari |
35 |
narrow for a centroid of high entropy where the values are evenly dispersed. This measure therefore |
2 |
26 Feb 07 |
jari |
36 |
may select centroids that don't have large expression variability in terms of range of values and |
2 |
26 Feb 07 |
jari |
37 |
hence appear rather flat across measurements. |
2 |
26 Feb 07 |
jari |
38 |
<br><br> |
2 |
26 Feb 07 |
jari |
39 |
</basefont> |
2 |
26 Feb 07 |
jari |
40 |
</body> |
2 |
26 Feb 07 |
jari |
41 |
</html> |