Compute scores for each cluster — scoreClusters • decorate

For each cluster compute summary statistics for the cluster to measure how strong the correlation structure is. Clusters with weak correlation structure can be dropped from downstream analysis.

Usage

scoreClusters(treeList, treeListClusters, BPPARAM = SerialParam())

Arguments

treeList: list of hclust objects
treeListClusters: from createClusters()
BPPARAM: parameters for parallel evaluation

Value

for all pairs of peaks within windowSize, report distance

Details

For each cluster, extract the correlation matrix and return the mean absolute correlation; the 75th, 90th and 95th quantile absolute correlation, and LEF, the leading eigen-value fraction which is the fraction of variance explained by the leading eigen value of the matrix abs(C).

Examples

library(GenomicRanges)
library(BiocParallel)

data('decorateData')

# Evaluate hierarchical clustering
treeList = runOrderedClusteringGenome( simData, simLocation ) 
#> 
Evaluating:chr20          
#> 

# Choose cutoffs and return clusters
treeListClusters = createClusters( treeList )
#> Method:capushe

# Evaluate score for each cluster
clstScore = scoreClusters(treeList, treeListClusters, BPPARAM = SerialParam() )
#> Evaluating strength of each cluster...
#> 
#> Dividing work into 1 chunks...