runPermutedData: Run hierarchical clustering permuting features

In GabrielHoffman/decorate: Differential Epigenetic Coregulation Test

Run hierarchical clustering permuting features

Description

Run hierarchical clustering permuting features to get statistics under the null

Usage

runPermutedData(
  X,
  gr,
  method = c("adjclust", "hclustgeo"),
  quiet = FALSE,
  alpha = 0.5,
  adjacentCount = 500,
  setNANtoZero = FALSE,
  method.corr = c("pearson", "spearman"),
  meanClusterSize = c(5, 10)
)

Arguments

X	data matrix were *rows* are features in sequential order
gr	GenomicRanges object with entries corresponding to the *rows* of X
method	'adjclust': adjacency constrained clustering. 'hclustgeo': incorporate data correlation and distance in bp
quiet	suppress messages
alpha	use by 'hclustgeo': mixture parameter weighing correlations (alpha=0) versus chromosome distances (alpha=1)
adjacentCount	used by 'adjclust': number of adjacent entries to compute correlation against
setNANtoZero	replace NAN correlation values with a zero
method.corr	Specify type of correlation: "pearson", "kendall", "spearman"
meanClusterSize	select target mean cluster size. Can be an array of values

Value

list of clusterScores and cutoff values at 5

Examples

library(GenomicRanges)

# load data
data('decorateData')

# First, analysis of original data
# Evaluate hierarchical clustering
treeList = runOrderedClusteringGenome( simData, simLocation ) 

# Choose cutoffs and return clusters
treeListClusters = createClusters( treeList, method='meanClusterSize', meanClusterSize=c(5, 10) )

# Evaluate score for each cluster
clstScore = scoreClusters(treeList, treeListClusters )

# Then, analysis of permuted data
# Evaluate hierarchical clustering
res = runPermutedData( simData, simLocation, meanClusterSize=c(5, 10)  ) 

# LEF values for permuted data at 5% false positive rate
res$cutoffs$LEF

# Retain clusters that pass this criteria
clustInclude = retainClusters( clstScore, "LEF", res$cutoffs$LEF )

GabrielHoffman/decorate documentation built on May 23, 2023, 1:29 a.m.