findClusters: Identification of clusters with similar temporal regulation

In Biopeak: Identification of Impulse-Like Gene Expression Changes in Short Genomic Series Data

Description

The findClusters function estimates the number of genes with similar temporal regulation and supports three different clustering algorithms: kmeans, dbscan and hierarchical clustering. Clustering is based on a PCA projection of the input data.

Usage

findClusters(peakdet, exprmat, maxclusters = 3, eps = 0.02,
  clusters = 3, method = "kmeans")

Arguments

peakdet	A list returned by the peakDetection function.
exprmat	A numeric matrix with expression series data with variables as rownames.
maxclusters	Maximal number of clusters used for kmeans cluster estimation.
eps	Epsilon value used by the dbscan algorithm.
clusters	Number of clusters used for the cutree function of the hierarchical clustering.
method	A character string defining the clustering algorithm with options: c('kmeans', 'dbscan', 'hclust').

Value

Returns a cluster assignment of each variable and the number of identified clusters.

Author(s)

David Lauenstein

Examples

# Example based on the heat-shock dataset
data(heat)
heat = as.matrix(heat)
# Define series
series <- c(37,40,41,42,43)
# Run the peak detection algorithm
peakdet <- peakDetection(heat, series, type ='rnaseq', actstrength = 1.5,
prominence = 1.3, minexpr = 5000)
# Cluster exploration using kmeans with a maximum of 4 clusters to be assigned
clusters <- findClusters(peakdet, heat, maxclusters = 4, method = 'kmeans')

Biopeak documentation built on Aug. 21, 2019, 5:10 p.m.