Mutual information network (KDE)
For continuous data (e.g. AQUA), mutual information is calculated between each pairwise combination of markers using variable, per-marker bandwidth kernel density estimation (KDE). Statistical significance is determined using 100,000 score permutations and multiple hypothesis correction is performed according to either Benjamini-Yekutieli FDR [Benjamini & Yekutieli 2001 Ann. Statist. 29:1165] (default) or Bonferroni to give a corrected p-value. The result is displayed as a network (a plot of markers connected pairwise by edges/lines) with a default significance threshold of FDR p≤0.05. The threshold value can subsequently be adjusted interactively. Mutual information provides an information-theoretic measure of interdependence between marker scores and can detect both linear and non-linear relationships.
Available for: continuous scoring
[Top]Viewing the results
See accessing and interpreting network results.
[Top]Example output

Figure 1: An example protein network with edges representing significant mutual information relationships