In aarkatkar/gelClust: Clustering and pathway analysis using exponential distributions

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

First, the gelClust library is loaded along with a dataset consisting of the TCDD-induced fold change in expression of 932 DEGs across 14 mouse strains and three traits: TCDD burden, change in fat percentage, and change in body weight gain.

library(gelClust)
data(foldChange)
data(traitData)

Then, the number of clusters for signed gamma clustering is selected through maximization of the log-likelihood function. This is chosen in a range from 2 up to the number of samples.

gm <- gelMatrix(foldChange)
score <- gammaTest(foldChange, gm, rbound=14)

As observed, the optimal number of clusters is 7. We can view these clusters with the gammaPlot function.

m <- gelSVD(foldChange, 7)
clust <- gammaCluster(gm, m)

set.seed(2)
gammaPlot(gm, clust$labels)

Within each of these clusters, we can further obtain sub-clusters such that genes from distinct sub-clusters within the same primary cluster are uncorrelated after controlling for the cluster itself.

sub.clust <- subClusters(foldChange, clust, 10)
table(sub.clust$labels)

aarkatkar/gelClust documentation built on May 6, 2022, 12:24 a.m.