Type: Package
Title: Order Restricted Clustering for Microarray Experiments
Version: 2.0.2
Date: 2015-07-15
Author: Adetayo Kasim, Martin Otava, Tobias Verbeke
Maintainer: Rudradev Sengupta <rudradev.sengupta@uhasselt.be>
Description: Provides clustering of genes with similar dose response (or time course) profiles. It implements the method described by Lin et al. (2012).
Imports: Iso
License: GPL-3
LazyLoad: yes
Repository: CRAN
Repository/R-Forge/Project: orcme
Repository/R-Forge/Revision: 65
Repository/R-Forge/DateTimeStamp: 2015-07-23 12:31:52
Date/Publication: 2015-07-31 12:12:01
NeedsCompilation: no
Packaged: 2015-07-27 14:12:53 UTC; rforge
Depends: R (≥ 2.10)

Order restricted clustering for dose-response trends in microarray experiments

Description

The function performs delta-clustering of a microarray data. It can be used for clustering of both the time-course or dose-response microarray data.

Usage

ORCME(DRdata, lambda, phi, robust=FALSE)

Arguments

DRdata

matrix of a microarray data with rows corresponding to genes and columns corresponding to time points or different doses

lambda

assumed proportion of coherence relative to the observed data, it ranges between 0 and 1. A lambda value of 1 considers the observed data as a cluster and lambda value of 0 finds every possible pattern within the data.

phi

minimum number of genes in a cluster

robust

logical variable that determines, if algorithm uses robust version based on median polish and absolute values, instead of mean square error. Default is FALSE.

Value

The matrix of classification into clusters: each row represents one gene and columns found clusters. The matrix consist of the Booleans values, in each row there is only one of them TRUE which means that the gene was classified into the respective cluster.

Author(s)

Adetayo Kasim, Martin Otava and Tobias Verbeke

References

Lin D., Shkedy Z., Yekutieli D., Amaratunga D., and Bijnens, L. (editors). (2012) Modeling Dose-response Microarray Data in EarlyDrug Development Experiments Using R. Springer.

Cheng, Y. and Church, G. M. (2000). Biclustering of expression data. In: Proceedings of the Eighth International Conference on Intelligent Systems for Molecular Biology, 1, 93-103.

See Also

monotoneDirection, plotIsomeans

Examples

  data(doseData)
  data(geneData)

  dirData <- monotoneDirection(geneData = geneData,doseData = doseData)
  incData <- as.data.frame(dirData$incData)
  
  
  print(orcme <- ORCME(DRdata=incData,lambda=0.15,phi=2))
  orcmeRobust <- ORCME(DRdata=incData,lambda=0.15,phi=2, robust=TRUE)
  
  # number of genes within cluster
  colSums(orcme)
  colSums(orcmeRobust)

Dose Data Example

Description

Dose data; a vector of length 12 with 3 observations for each of 4 doses.

Usage

data(doseData)

Format

The format is: num [1:12] 1 1 1 2 2 2 3 3 ...

Examples


data(doseData)
  
doseData

Gene Expression Data Example

Description

This dose-response microarray data contains 1000 genes and 4 doses (one control dose (zero dose) and three increasing dose) with 3 arrays at each dose level. Due to confidetiality, it is only part of the real data set.

Usage

data(geneData)

Format

A data frame with 1000 observations on the following 12 variables.

X1

Sample one with zero dose

X1.1

Sample two with zero dose

X1.2

Sample three with zero dose

X2

Sample one with second dose

X2.1

Sample two with second dose

X2.2

Sample three with second dose

X3

Sample one with third dose

X3.1

Sample two with third dose

X3.2

Sample three with third dose

X4

Sample one with fourth dose

X4.1

Sample two with fourth dose

X4.2

Sample three with fourth dose

References

Testing for Trend in Dose-Response Microarray Experiments: a Comparison of Testing Procedures, Multiplicity, and Resampling-Based Inference, Lin et al. 2007, Stat. App. in Gen. & Mol. Bio., 6(1), article 26.

Examples

  
data(geneData)
  
head(geneData)

The monotone means under increasing/decreasing trend

Description

The function calculates the likelihood for the increasing and decreasing trend in the dose response for all the given genes separately gene-by-gene. The trend with the higher likelihood is chosen and the isotonic regression is applied on the means.

Usage

monotoneDirection(geneData, doseData)

Arguments

geneData

gene expression matrix for all genes

doseData

indicates the dose levels

Value

A list with components

direction

the direction with the higher likelihood of increasing (indicated by "up") or decreasing (indicated by "dn") trend.

incData

isotonic means with respect to dose for those genes that were classified as following the increasing trend.

decData

isotonic means with respect to dose for those genes that were classified as following the decreasing trend.

obsincData

observed gene expression matrix for those genes that were classified as following the increasing trend.

obsdecData

observed gene expression matrix for those genes that were classified as following the decreasing trend.

arrayMean

isotonic means with respect to dose for all genes.

Author(s)

Adetayo Kasim, Martin Otava and Tobias Verbeke

References

Lin D., Shkedy Z., Yekutieli D., Amaratunga D., and Bijnens, L. (editors). (2012) Modeling Dose-response Microarray Data in Early Drug Development Experiments Using R. Springer.

Cheng, Y. and Church, G. M. (2000). Biclustering of expression data. In: Proceedings of the Eighth International Conference on Intelligent Systems for Molecular Biology, 1, 93-103.

See Also

ORCME, plotIsomeans

Examples

  data(doseData)
  data(geneData)

  dirData <- monotoneDirection(geneData = geneData,doseData = doseData)

  ## direction of monotone trend
  Direction <- dirData$direction
  ## Isotonic means for upward genes
  incData <- as.data.frame(dirData$incData)
  ##Isotonic means for downward genes
  decData <- as.data.frame(dirData$decData)
  ## observd data upward genes
  obsIncData <- as.data.frame(dirData$obsincData)
  ## observed data for downward genes
  obsDecData <- as.data.frame(dirData$obsdecData)
  ## isotonic means for all genes
  isoMeans <- as.data.frame(dirData$arrayMean)

Plotting the gene specific profiles for one given cluster of genes

Description

The function is plotting the profiles of the genes that belongs to the same cluster. It is not providing the clustering itself, just plotting the results of clustering from input. Optionally, the function can center the profiles around the gene-specific means.

Usage

plotCluster(DRdata, doseData, ORCMEoutput, clusterID, 
zeroMean=FALSE, xlabel, ylabel, main="")

Arguments

DRdata

the microarray data with rows corresponding to genes and columns corresponding to time points or different doses

doseData

indicates the dose levels

ORCMEoutput

the matrix of classification into clusters: each row represents one gene and columns found clusters. The matrix consist of the Booleans values, in each row there is only one of them TRUE which means that the gene was classified into the respective gene

clusterID

id of the cluster to be plotted

zeroMean

if TRUE, it centers the gene profiles around the gene-specific means, default is FALSE

xlabel

a title for the x axis

ylabel

a title for the y axis

main

an overall title for the plot

Value

Plot of the gene specific profiles dependent one the dose level (or time point) that are classified into the given cluster.

Author(s)

Adetayo Kasim, Martin Otava and Tobias Verbeke

References

Lin D., Shkedy Z., Yekutieli D., Amaratunga D., and Bijnens, L. (editors). (2012) Modeling Dose-response Microarray Data in Early Drug Development Experiments Using R. Springer.

Cheng, Y. and Church, G. M. (2000). Biclustering of expression data. In: Proceedings of the Eighth International Conference on Intelligent Systems for Molecular Biology, 1, 93-103.

See Also

ORCME, plotIsomeans

Examples

  data(doseData)
  data(geneData)

  
  dirData <- monotoneDirection(geneData = geneData,doseData = doseData)
  incData <- as.data.frame(dirData$incData)
  ORCMEoutput <- ORCME(DRdata=incData,lambda=0.15,phi=2)
  
  plotCluster(DRdata=incData,doseData=doseData, ORCMEoutput=ORCMEoutput,
  clusterID=4,zeroMean=FALSE, xlabel="Dose",ylabel="Gene Expression")

Plot of the observed gene expression and the isotonic means with respect to dose

Description

The function is plotting the observed data points of the gene expression and isotonic means with respect to dose for one particular gene.

Usage

plotIsomeans(monoData, obsData, doseData, geneIndex)

Arguments

monoData

isotonic means with respect to dose for all genes

obsData

observed gene expression for all genes

doseData

indicates the dose levels

geneIndex

index of the gene to be plotted

Value

Plot of the data points and the isotonic means for each dose with the isotonic regression curve.

Author(s)

Adetayo Kasim, Martin Otava and Tobias Verbeke

References

Lin D., Shkedy Z., Yekutieli D., Amaratunga D., and Bijnens, L. (editors). (2012) Modeling Dose-response Microarray Data in Early Drug Development Experiments Using R. Springer.

Cheng, Y. and Church, G. M. (2000). Biclustering of expression data. In: Proceedings of the Eighth International Conference on Intelligent Systems for Molecular Biology, 1, 93-103.

See Also

ORCME, monotoneDirection

Examples

  data(doseData)
  data(geneData)

  dirData <- monotoneDirection(geneData = geneData,doseData = doseData)
  incData <- as.data.frame(dirData$incData)
  obsIncData <- as.data.frame(dirData$obsincData)
  
  ## gene-specific profile plot
  plotIsomeans(monoData=incData,obsData=obsIncData,doseData=
  doseData,geneIndex=10)

Plot the variaty of the properties dependent on the proportion of heterogeneity in observed data set

Description

This function provides the plots of the dependency of the variety of properties on the proportion of heterogeneity in observed data set. It is not using the clustering as simple input, but it is also computing additional properties. The function can plot within cluster sum of squares, number of cluster, penalized within cluster sum of squares, Calsanzik and Harabasx index and Hartigan index.

Usage

  plotLambda(lambdaChoiceOutput,output)

Arguments

lambdaChoiceOutput

the output of the function resampleORCME

output

the variable that determines which output would be plotted, the values are "wss" for the cluster sum of squares, "ncluster" for the number of cluster, "pwss" for the penalized within cluster sum of squares, "ch" for the Calsanzik and Harabasx index and "h" for the Hartigan index

Value

A plot of one of the properties mentioned above dependent on the proportion of heterogeneity. The confidence intervals are plotted instead of the point estimates.

Author(s)

Adetayo Kasim, Martin Otava and Tobias Verbeke

References

Lin D., Shkedy Z., Yekutieli D., Amaratunga D., and Bijnens, L. (editors). (2012) Modeling Dose-response Microarray Data in Early Drug Development Experiments Using R. Springer.

Cheng, Y. and Church, G. M. (2000). Biclustering of expression data. In: Proceedings of the Eighth International Conference on Intelligent Systems for Molecular Biology, 1, 93-103.

See Also

ORCME, resampleORCME

Examples

  data(doseData)
  data(geneData)

  dirData <- monotoneDirection(geneData = geneData,doseData = doseData)
  incData <- as.data.frame(dirData$incData)
 
  lambdaVector <- c(0.05,0.50,0.95)
  
  
  lambdaChoiceOutput <- resampleORCME(clusteringData=incData, lambdaVector=lambdaVector)
  plotLambda(lambdaChoiceOutput,output="wss")
  plotLambda(lambdaChoiceOutput,output="ncluster")
  plotLambda(lambdaChoiceOutput,output="pwss")
  plotLambda(lambdaChoiceOutput,output="ch")
  plotLambda(lambdaChoiceOutput,output="h")

Estimation of the proportion of the heterogeneity in the observed data for clustering

Description

The function is computing within cluster sum of squares for given proportion of heterogeneity. Minimal number of genes per cluster is fixed as 2. The sum of squares is computed through resampling the 100 data sets with 100 genes randomly sampled with replacement from the reduced expression data.

Usage

resampleORCME(clusteringData, lambdaVector, robust=FALSE)

Arguments

clusteringData

the microarray data with rows corresponding to genes and columns corresponding to time points or different doses

lambdaVector

vector of assumed proportions of of heterogeneity of the observed data, it ranges between 0 and 1. A lambda value of 1 considers the observed data as a cluster and lambda value of 0 finds every possible pattern within the data

robust

logical variable that determines, if algorithm uses robust version based on median polish and absolute values, instead of mean square error. Default is FALSE.

Value

A list of matrices that represent one of the 100 iterations. Every matrix consist of the columns

lambda

vector of the proportions of heterogeneity given as input

WSS

within clusters sum of squares for given proportion of heterogeneity

TSS

total clusters sum of squares for given proportions of heterogeneity

nc

number of clusters as a function for given proportions of heterogeneity

Author(s)

Adetayo Kasim, Martin Otava and Tobias Verbeke

References

Lin D., Shkedy Z., Yekutieli D., Amaratunga D., and Bijnens, L. (editors). (2012) Modeling Dose-response Microarray Data in Early Drug Development Experiments Using R. Springer.

Cheng, Y. and Church, G. M. (2000). Biclustering of expression data. In: Proceedings of the Eighth International Conference on Intelligent Systems for Molecular Biology, 1, 93-103.

See Also

ORCME, plotLambda

Examples

  data(doseData)
  data(geneData)

  dirData <- monotoneDirection(geneData = geneData,doseData = doseData)
  incData <- as.data.frame(dirData$incData)
 
  lambdaVector <- c(0.05,0.50,0.95)
  
  
  resampleORCME(clusteringData=incData, lambdaVector=lambdaVector, robust=FALSE)