rhcoclust {rhcoclust}

The function for co-clustering sample and feature to explore significant samples and their regulatory features

Description

Toxicogenomic studies require co-clustering to identify biomarker genes for the assessment of chemical toxicity from gene expression levels. It is also essential in the drug discovery experiments. However, gene expression datasets are often contaminated by outliers due to several steps involve in the data generating process. This package performs robust hierarchical co-clustering between row and column entities of a data matrix by reducing the influence of outlying observations. It can be used to explore biomarker genes those are divided into upregulatory and downregulatory groups by the influence of different chemical compounds groups more accurately. It can also provide the statistical significance of the identified co-clusters.

Usage

rhcoclust(data, rk, ck, method.dist = "manhattan", method.hclust= "ward.D")

Arguments

data

A data matrix containing data having the characteristics of interval and ratio level of measurement or continuous data

rk

Number of clusters in the row entities of the data matrix

ck

Number of clusters in the column entities of the data matrix

method.dist

The distance measure to be used. The default is "manhattan". The other options are "euclidean", "maximum", "canberra", "binary" or "minkowski". Any unambiguous substring can be given.

method.hclust

The agglomeration method to be used. The default is "ward.D". The other options are "ward.D2", "single", "complete", "average" (= UPGMA), "mcquitty" (= WPGMA), "median" (= WPGMC) or "centroid" (= UPGMC).

Value

A list of object that containing the following:

Coclust_MeanMat : A data frame containing combination of row and column cluster number in the first column and their ranked co-cluster mean in the second column. In the first column first number indicates row cluster index and second number indicates column cluster index, respectively.

CoClsDtMat : The reorganized transformed data matrix to generate co-cluster graph.

NG_Cocls : The index of gene/row names.

NC_Cocls : The index of column names.

rowclust :The gene/row entity clusters.

colclust : The column entity clusters.

colorsG: Colors of genes/row entity clusters to generate co-cluster graph.

colorsC: Colors of DCCs/column entity clusters to generate co-cluster graph.

CentralLine: Central Line of individual control chart to generate graph of control chart and to identify significant co-clusters.

UpContLimit: Upper Control Limit to generate graph of control chart and to identify significant co-clusters.

LowrContLimit: Lower Control Limit to generate graph of control chart and to identify significant co-clusters.

color: Colors to generate individual control chart.

pchmark: Shape of points to generate individual control chart.

Author(s)

Md. Bahadur Badsha <[e-mail]>

Examples

# Load necessary library
library(rhcoclust)
library(fields)

# Load real data
data("FCGE_Data_GMP")
data("FCGE_Data_PPARs")
# Load predefined real data
# Real data use: data <- FCGE_Data_PPARs
# Real data use: data <- FCGE_Data_GMP

# Load predefined simulated data
data("simu_data")

# simulated data
data <- simu_data
# Apply rhcoclust to identify significant co-cluster of samples and their regulatory features
CoClustObj <- rhcoclust(data, rk=4, ck=3, method.dist = "manhattan", method.hclust = "ward.D")
# For real data either FCGE_Data_PPARs or FCGE_Data_GMP
#CoClustObj <- rhcoclust(data, rk=3, ck=3, method.dist = "manhattan", method.hclust = "ward.D")

# A data frame containing combination of row and column cluster number in the first
# column and their ranked co-cluster mean in the second cluster.
#GC_cls_MeanMat <- CoClustObj$Coclust_MeanMat

# The reorganized transformed data matrix to generate co-cluster graph.
CoClsDtMat <- CoClustObj$CoClsDtMat

# The gene/row entity clusters.
rowclust <- CoClustObj$rowclust

# The column entity clusters.
colclust <- CoClustObj$colclust

# Colors of genes/row entity clusters to generate co-cluster graph
colorsG <- CoClustObj$colorsG

# Colors of DCCs/column entity clusters to generate co-cluster graph
colorsC <- CoClustObj$colorsC

# Central Line of individual control chart to generate graph of control chart and to
# identify significant co-clusters.
CntrLine_QC <- CoClustObj$CentralLine

# Upper Control Limit to generate graph of control chart and to identify significant
# co-clusters.
UCL_QC <- CoClustObj$UpContLimit

# Lower Control Limit to generate graph of control chart and to identify significant
# co-clusters.
LCL_QC <- CoClustObj$LowrContLimit

# Colors to generate individual control chart.
ColorQC <- CoClustObj$color

# Shape of points to generate individual control chart.
PcmQC <- CoClustObj$pchmark

# Plot co-cluster
# par(mar=c(6,10,3,6)) # Modify if needed
# mar order: bottom, left, top, and right
# please use different values if needed for cex.xaxis and cex.yaxis
# to adjust xaxis and yaxis text
plot_rhcoclust (CoClustObj, plot.coclust = TRUE, plot.SCC = FALSE)

# Plot SCC
# use dev.off() to avoid the figure margin from previous plot
plot_rhcoclust (CoClustObj, plot.coclust = FALSE, plot.SCC = TRUE)