Filtering of gene sets

Description

Filters gene sets according to different criteria.

Usage

filterGeneSets(
	geneSets,
	includedGenes = NULL,
	minIntersectSize = length(includedGenes),
	adjMatrix = NULL,
	steps = NULL)

Arguments

Value

Returns a filtered list of gene sets with the same structure as geneSets.

See Also

geneSetAnalysis, preprocessGs

Examples

geneSets <- list(
		gs1 = paste("gene",1:20,sep=""),
		gs2 = paste("gene",50:60,sep=""),
		gs3 = paste("gene",90:92,sep=""),
		gs4 = paste("gene",55:65,sep="")
	)

newGeneSets1 <- filterGeneSets(
	geneSets = geneSets,
	includedGenes = c("gene55","gene60"))

newGeneSets2 <- filterGeneSets(
	geneSets = geneSets,
	includedGenes = c("gene1","gene55","gene20","gene100"),
	minIntersectSize = 2)

examplePathway <- c("gene1","gene2","gene3","gene4")
pathwayAdjMatrix <- matrix(0,100,100)
rownames(pathwayAdjMatrix) <- paste("gene",1:100,sep="")
colnames(pathwayAdjMatrix) <- paste("gene",1:100,sep="")

# gene1 interacts with gene2 and gene3
# -> step 1 if gene1 is starting point
pathwayAdjMatrix[1,2:3] <- 1
pathwayAdjMatrix[2:3,1] <- 1

# gene3 interacts with gene4
# -> step 2 if gene1 is starting point
pathwayAdjMatrix[3,4] <- 1
pathwayAdjMatrix[4,3] <- 1

newGeneSets3 <- filterGeneSets(
	geneSets = geneSets,
	includedGenes = "gene1",
	adjMatrix = pathwayAdjMatrix,
	steps = 2)

Main interface for enrichment analyses.

Description

The main function of the package that performs a gene set analysis for a list of gene sets.

Usage

geneSetAnalysis(
	...,
	dat,
	geneSets,
	analysis,
	signLevel = 0.05,
	preprocessGeneSets = FALSE,
	adjustmentMethod = p.adjust.methods,
	cluster = NULL)

Arguments

Details

This is the main interface function of the package for gene set enrichment analyses. Analyses usually consist of a pipeline of steps. Often, the first step is the calculation of a summary statistic for the relation of each gene to the class labels. These values or transformations thereof are employed to calculate a gene set statistic for each of the supplied gene sets. The significance of gene set enrichments can be determined according to different methods, and the robustness of gene sets can be evaluated by slightly modifying the gene sets. To provide a flexible mechanism for the plethora of different approaches arising from the different choices, basic pipeline configurations are encapsulated in gsAnalysis objects which can be created using the gsAnalysis function. Ready-to-use configuration objects for certain well-known methods are included in the package (see predefinedAnalyses). Parameters of the chosen analysis pipeline can be set in the ... parameter.

Value

An object of the type gsaResult with the following elements:

References

Ackermann, M., Strimmer, K. (2009) A general modular framework for gene set enrichment analysis. BMC Bioinformatics, 10(1), 47.

See Also

gsAnalysis, gls, transformation, gss, global, significance, evaluateGeneSetUncertainty, hist.gsaResult, preprocessGs

Examples

# load data
data(exampleData)
  
  # apply predefined analysis for gene set enrichment analysis
  res <- geneSetAnalysis(	
  	# parameters for geneSetAnalysis
  	dat = countdata,
  	geneSets = pathways[1],
  	analysis = analysis.averageCorrelation(),
  	adjustmentMethod = "fdr",
  	# additional parameters for analysis.averageCorrelation
  	labs = labels,
  	method = "pearson",
  	numSamples = 10)

Gene-level statistics

Description

Functions to calculate the gene-level statistic, as used in the gls parameter of gsAnalysis. A gene-level statistic calculates a measure of correlation between the expression of a gene and the class labels.

Usage

gls.cor(dat, labs, method = "pearson")

gls.regression(dat, labs)

gls.foldChange(dat, labs, logMeasurements = TRUE)

gls.tStatistic(dat, labs, pValue = FALSE, alternative = "two.sided")

gls.moderateTStatistic(dat,labs)

gls.nBinomTest(dat, labs,
	returnValue = c("pval", "qval", "foldChange", "log2FoldChange"),
	dispersionMethod = "blind",
	dispersionSharingMode = "fit-only",
	dispersionFitType = "local")

Arguments

Details

Standard functions for the calculation of gene-level statistics (to be used in an analysis pipeline defined by gsAnalysis):

• gls.cor: Calculates the correlation of the gene expression values to the class labels.

• gls.regression: Calculates the slope of a linear regression of the gene expression values and the class labels.

• gls.foldChange: Calculates the (standard or log2) fold change between the measurements for the two classes.

• gls.tStatistic: Calculates the p-value or the statistic of a two-sample t test for the measurements of the two classes

• gls.moderateTStatistic: Calculates the moderate t statistic for the measurements of the two classes

• gls.nBinomTest: Applies the negative binomial test for sequencing data based on the DESeq2 package to test for differences between two classes (see results).

Value

Each of these function returns a numeric vector of gene-level statistics (one entry per gene).

See Also

geneSetAnalysis, gsAnalysis, gss, transformation

Gene set statistics

Description

Functions to calculate a gene set statistic, as used in the gss parameter of gsAnalysis. A gene set statistic summarizes a single gene set.

Usage

gss.mean(x, geneSetIndices)

gss.sum(x, geneSetIndices)

gss.wilcoxonRankTest(x, geneSetIndices)

gss.maxmean(x, geneSetIndices)

gss.median(x, geneSetIndices)

gss.enrichmentScore(x, geneSetIndices, p = 1)

gss.fisherExactTest(x, geneSetIndices)

gss.gsz(x, geneSetIndices, w1 = 0.2, w2 = 0.5, preVar = 0, varConstant = 10)

Arguments

Details

Standard functions for the calculation of gene set statistics (to be used in an analysis pipeline defined by gsAnalysis):

• gss.mean: Calculates the mean of the (transformed) gene-level statistic values for the genes in the set.

• gss.sum: Calculates the sum of the (transformed) gene-level statistic values for the genes in the set.

• gss.wilcoxonRankTest: Calculates a Wilcoxon test comparing the (transformed) gene-level statistic values for the genes in the set versus those of the genes not in the set.

• gss.maxmean: Calculates the maximum of the means of positive and negative statistic values, weighted by the overall proportion of positive/negative values (e.g. for correlation scores where the sign denotes the direction). Described in Efron and Tibshirani.

• gss.median: Calculates the median of the (transformed) gene-level statistic values for the genes in the set.

• gss.enrichmentScore: Calculates the enrichment score of the (transformed) gene-level statistic values for the genes in the set, as described in Subramanian et al.

• gss.fisherExactTest: Performs Fisher's exact test to check gene sets for overrepresentation in the differential genes. This should be used used in combination with the transformation transformation.adjustAndBinarize.

• gss.gsz: Calculates the Gene Set Z-score of the (transformed) gene-level statistic values for the genes in the set. Described in Toronen et al.

Value

Each method returns a single numeric value, the gene set statistic for the supplied gene set.

References

Efron, B., Tibshirani, R. (2007) On testing the significance of sets of genes. Annals of Applied Statistics, 1, 107-129.

Subramanian, A., Tamayo, P., Mootha, V. K., Mukherjee, S., Ebert, B. L., Gillette, M. A., Paulovich, A., Pomeroy, S. L., Golub, T. R., Lander, E. S., Mesirov, J. P. (2005) Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Science of the United States of America, 102, 15545–15550.

Toronen, P., Ojala, P. J., Marttinen, P., Holm L. (2009) Robust extraction of functional signals from gene set analysis using a generalized threshold free scoring function. BMC Bioinformatics, 10(1), 307.

See Also

geneSetAnalysis, gsAnalysis, gls, transformation

Enrichment analysis

Description

Toolbox for gene set analysis of uncertain gene sets.

Details

This package provides an approach for evaluating the fuzziness of a gene set. This is done by repeatedly performing gene set analyses on slightly modified versions of the gene set and comparing their enrichment scores. A utility for such uncertainty tests is provided in the evaluateGeneSetUncertainty function.

The package also comprises a generic framework for different types of enrichment analyses (Ackermann and Strimmer). It establishes a customizeable pipeline that typically consists of the following steps:

• Calculation of gene-level statistics: A gene-level statistic scores the relationship between the measurements for a specific gene and the class labels. Typical measures include correlation coefficients, the t statistic or the fold change between the groups (see gls for gene-level statistics included in the package).

• Transformation of gene-level statistic values: Optionally, the gene-level statistic values can be postprocessed, e.g. by taking the absolute value or the square for correlation values or by binarizing or ranking values. See transformation for transformations included in the package.

• Calculation of gene set statistics: Based on the (possibly transformed) gene-level statistics, the gene set(s) of interest is/are scored. Examples are the median, the mean or the enrichment score of the gene-level statistic values in the gene set(s). See gss for gene set statistics included in the package.

• Significance assessment: To assess the significance of the gene set statistic value(s) with respect to a null distribution, computer-intensive tests are performed. These tests repeatedly sample random label vectors or gene sets and calculate their gene set statistic values. These values can then be compared to the true gene set statistics. See significance for significance assessment methods included in the package.

The package represents such analysis pipelines as configuration objects that can be created using the function gsAnalysis. For predefined state-of-the-art methods, such as Gene Set Enrichment Analysis (Subramanian et al), Overrepresentation Analysis or Global Ancova (Hummel et al), it provides predefined configurations (see predefinedAnalyses).

The main function for standard gene set analyses, geneSetAnalysis, performs enrichment analyses based on pipeline configuration objects.

Author(s)

Florian Schmid, Christoph Müssel, Johann M. Kraus, Hans A. Kestler

Maintainer: Hans A. Kestler <hans.kestler@uni-ulm.de>

References

Ackermann, M., Strimmer, K. (2009) A general modular framework for gene set enrichment analysis. BMC Bioinformatics, 10(1), 47.

Subramanian, A., Tamayo, P., Mootha, V. K., Mukherjee, S., Ebert, B. L., Gillette, M. A., Paulovich, A., Pomeroy, S. L., Golub, T. R., Lander, E. S., Mesirov, J. P. (2005) Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Science of the United States of America, 102, 15545-15550.

Hummel, M., Meister, R., Mansmann, U. (2008) Globalancova: exploration and assessment of gene group effects. Bioinformatics, 24(1), 78–85.

Examples

  data(exampleData)
  ##################################
  # Example 1: gene set analysis   #
  ##################################
  res <- geneSetAnalysis(  
    # parameters for geneSetAnalysis
    dat = countdata,
    geneSets = pathways[1],
    analysis = analysis.averageCorrelation(),
    adjustmentMethod = "fdr",
    # additional parameters for analysis.averageCorrelation
    labs = labels,
    method = "pearson",
    numSamples = 50)
  
  summary(res, mode="table")
  
  ####################################
  # Example 2: uncertainty analysis  #
  ####################################
  resUncertainty <- evaluateGeneSetUncertainty(  
    # parameters for evaluateGeneSetUncertainty
    dat = countdata,
    geneSet = pathways[[3]],
    analysis = analysis.averageCorrelation(),
    numSamplesUncertainty = 5,
    blockSize = 1,
    k = seq(0.1,0.9,by=0.1),
    # additional parameters for analysis.averageCorrelation
    labs = labels,
    numSamples = 5)
  
  plot(resUncertainty, main = names(pathways[3]))

Global analyses

Description

Functions to perform global gene set analyses, as used in the globalStat parameter of gsAnalysis.

Usage

global.overrepresentation(dat,
	geneSet,
	coreSet)

global.ancova(dat,
	geneSet,
	labs,
	...)

global.test(dat,
	geneSet,
	labs,
	...)

Arguments

Details

Wrapper functions for global gene set analyses.

• global.overrepresentation: This function performs an overrepresentation analysis by rating the overlap of geneSet and coreSet with respect to the set of all genes using Fisher's exact test.

• global.ancova: This function performs a global gene set enrichment analysis using the global ANCOVA method by Hummel et al. It wraps the GlobalAncova function in the GlobalAncova package.

• global.test: This function performs a global gene set enrichment analysis using a global test by Goeman et al. It wraps the gt function in the globaltest package.

Value

A list containing the following items:

References

Hummel, M., Meister, R., Mansmann, U. (2008) GlobalANCOVA: exploration and assessment of gene group effects. Bioinformatics, 24(1), 78–85.

Goeman, J. J., van de Geer, S. A., de Kort, F., van Houwelingen, H. C. (2004) A global test for groups of genes: testing association with a clinical outcome. Bioinformatics, 20(1), 93–99.

See Also

geneSetAnalysis, gsAnalysis

Significance assessment

Description

Functions to assess the significance of the gene-level statistics, as used in the significance parameter of gsAnalysis. These functions are based on applying the same analysis to randomly modified data sets or gene sets and comparing their statistic values to the original gene set statistic value.

Usage

significance.sampling(
	...,
	dat,
	geneSet,
	analysis,
	glsValues,
	numSamples = 1000)

significance.permutation(
	...,
	dat,
	geneSet,
	analysis,
	glsValues,
	numSamples = 1000,
	labs)

significance.restandardization(
	...,
	dat,
	geneSet,
	analysis,
	glsValues,
	numSamples = 1000,
	labs)

Arguments

Details

Standard methods for the significance assessment of a gene set statistic (to be used in an analysis pipeline defined by gsAnalysis):

• significance.sampling: This function repeatedly draws random gene sets. Their gene set statistic values form the null distribution.

• significance.permutation: This function repeatedly permutes the labels of the data set. The gene set statistic values for the original gene set on the permuted data set form the null distribution.

• significance.restandardization: This function applies both a gene set sampling and a label permutation. The permutation statistic values are standardized by their mean and standard deviation and then restandardized based on the gene set sampling statistic values. These restandardized values form the null distribution (Efron and Tibshirani).

Value

significance.sampling returns a list with the following elements:

significance.permutation returns a list with the following elements:

significance.restandardization returns a list with the following elements:

References

Efron, B., Tibshirani, R. (2007) On testing the significance of sets of genes. Annals of Applied Statistics, 1, 107-129.

See Also

geneSetAnalysis, gsAnalysis, hist.gsaResult

Transformations

Description

Functions to transform the gene-level statistic values prior to the calculation of the gene set statistics, as used in the transformation parameter of gsAnalysis. Most of the functions wrap existing R functions.

Usage

transformation.abs(x)

transformation.square(x)

transformation.localFdr(x,
	statistic="pvalue",
	cutoff.method="fndr",
	pct0=0.75)

transformation.binarize(x, quant)

transformation.rank(x)

transformation.adjust(x, adjMethod = "fdr")

transformation.adjustAndBinarize(x, adjMethod = "fdr", threshold = 0.05)

Arguments

Details

Standard transformation functions for gene-level statistics (to be used in an analysis pipeline defined by gsAnalysis):

• transformation.abs: Calculates the absolute values of the elements in x (a wrapper for abs).

• transformation.square: Squares all elements in x.

• transformation.localFdr: Calculates the local fdr for the elements in x. This is a wrapper for fdrtool.

• transformation.binarize: Binarizes the values in x by using the quant quantile as a threshold.

• transformation.rank: Ranks the values in x and returns the rank vector.

• transformation.adjust: Adjusts for multiple testing according to the adjustment method specified in adjMethod.

• transformation.adjustAndBinarize: Adjusts for multiple testing according to the adjustment method specified in adjMethod and binarizes the resulting p-values according to threshold (values smaller than the threshold become 1 others 0).

Value

All functions return a vector of transformed values having the same length as x.

See Also

geneSetAnalysis, gsAnalysis, gss, gls

Example data

Description

Randomly generated example data representing a count matrix

Usage

data(exampleData)

Details

The data consists of a variable countdata with a matrix with 1113 genes and 96 samples. The count values are randomly generated numbers.

Examples

data(exampleData)

print(countdata)

Create an overview table for an analysis

Description

Creates a data frame summarizing an analysis. This table has one row per gene set, each comprising the adjusted and unadjusted p-values and the number of genes for the set.

Usage

createSummaryTable(object,
		orderBy = c("adjustedPValues", "rawPValues", "geneSetName"),
		significantOnly = FALSE,
		signLevel = object$signLevel)

Arguments

Value

A data frame with one row for each included gene set and the columns "adjustedPValues", "rawPValues", "geneSetName" and "geneSetSize". For overrepresentation analyses, there is an additional column "intersectSize" specifying the size of the intersection of the core set and the corresponding gene set.

See Also

geneSetAnalysis, hist.gsaResult, summary

Examples

# load data
data(exampleData)
  # perform gene set analyses for several pathways
  res <- geneSetAnalysis(
  	# global parameters
	  dat = countdata,
	  geneSets = pathways, 
	  analysis = analysis.averageCorrelation(), 
	  # additional parameters for analysis.averageCorrelation
	  labs = labels,
	  numSamples = 10)

  tab <- createSummaryTable(res)

Quantify gene set uncertainty

Description

A robustness measure that quantifies the uncertainty of a gene set by performing a resampling experiment and can be used in the robustness parameter of gsAnalysis.

Usage

evaluateGeneSetUncertainty(
	...,
	dat,
	geneSet,
	analysis,
	numSamplesUncertainty,
	blockSize = 1,
	k = seq(0.01, 0.99, by=0.01),
	signLevel = 0.05,
	preprocessGeneSet = FALSE,
	cluster = NULL)

Arguments

Details

The uncertainty analysis repeatedly replaces parts of the original gene sets by random genes and calculating the gene set statistics for these randomized gene sets. This yields a distribution of gene set statistic values for slightly modified variants of the original gene set.

Value

Returns a list (of class uncertaintyResult) with the following elements:

See Also

geneSetAnalysis, gsAnalysis, gls, transformation, gss, plot.uncertaintyResult

Examples

data(exampleData)

  res <- evaluateGeneSetUncertainty(
  	# parameters for evaluateGeneSetUncertainty
  	dat = countdata,
  	geneSet = pathways[[1]],
  	analysis = analysis.averageCorrelation(),
  	numSamplesUncertainty = 10,
  	k = seq(0.1,0.9, by=0.1),
  	# additional parameters for analysis.averageCorrelation
  	labs = labels,
  	numSamples = 10)

Gene set analysis.

Description

Defines the configuration of an analysis that can be performed using geneSetAnalysis, and returns it as a wrapper object.

Usage

gsAnalysis(name,
	gls = NULL,
	glsParameterNames = NULL,
	transformation = NULL,
	transformationParameterNames = NULL,
	gss = NULL,
	gssParameterNames = NULL,
	globalStat = NULL,
	globalStatParameterNames = NULL,
	significance = NULL,
	significanceParameterNames = NULL,
	testAlternative = c("greater", "less"))

Arguments

Details

The function provides a way of flexibly defining the steps of the gene set analysis pipeline. This pipeline consists of a subset of the following steps, each of which may have specific parameters:

• Gene-level analysis: A gene-level statistic scores the relationship between the measurements for a specific gene and the class labels. Typical measures include correlation coefficients, the t statistic or the fold change between the groups (see gls for gene-level statistics included in the package).

• Transformation of gene-level statistics: Optionally, the gene-level statistic values can be postprocessed, e.g. by taking the absolute value or the square for correlation values or by binarizing or ranking values. See transformation for transformations included in the package.

• Gene set analysis: Based on the (possibly transformed) gene-level statistics, the gene set(s) of interest is/are scored. Examples are the median, the mean or the enrichment score of the gene-level statistic values in the gene set(s). See gss for gene set statistics included in the package.

• Significance assessment: To assess the significance of the gene set statistic value(s) with respect to a null distribution, computer-intensive tests are performed. These tests repeatedly sample random label vectors or gene sets and calculate their gene set statistic values. These values can then be compared to the true gene set statistics. See significance for significance assessment methods included in the package.

• Global analysis: As an alternative to the above pipeline steps, it is possible to define a single, global method that directly calculates an enrichment p-value for a supplied data set and gene set. See global for the global analysis tests included in the package.

Several state-of-the-art analyses have predefined configuration objects in which the above steps are defined accordingly (see predefinedAnalyses).

Value

An object of class gsAnalysis with components corresponding to the above parameters.

See Also

predefinedAnalyses, geneSetAnalysis, evaluateGeneSetUncertainty, gls, transformation, gss, global, significance

Examples

	data(exampleData)
# defines an analysis that corresponds to gsAna1()
gsa <- gsAnalysis(
	name = "averageCorrelation",
	gls = "gls.cor",
	glsParameterNames = c("labs","method"),
	transformation = "transformation.abs",
	transformationParameterNames = NULL,
	gss = "gss.mean",
	gssParameterNames = NULL,
	globalStat = NULL,
	globalStatParameterNames = NULL,
	significance = "significance.sampling",
	significanceParameterNames = c("numSamples"),
	testAlternative = "greater")
print(gsa)


  # apply the previously defined analysis
  res <- geneSetAnalysis(
  	# global parameters
  	dat = countdata,
  	geneSets = pathways[1], 
  	analysis = gsa, 
  	# parameters for the specific analysis gsAna1
  	labs = labels,
  	numSamples = 10)

Null distribution histogram and statistic of the input set for enrichment analyses.

Description

Plots the distribution of gene set statistic values obtained in different resampling settings of an enrichment analysis, and draws the statistic value of the input set as a vertical line.

Usage

## S3 method for class 'gsaResult'
hist(x,
	signLevel = x$signLevel,
	subset = NULL,
	ask = FALSE,
	addLegend = TRUE,
	...)

Arguments

Details

The function plots the distribution of gene set statistic values under the null hypothesis. It requires the significance assessment step of the enrichment analysis configuration (parameter significance or gsAnalysis) to be a computer-intensive testing procedure that yields a distribution of gene set statistic p-values under the null hypothesis. Predefined configurations for which this plot works are analysis.gsea, analysis.averageCorrelation and analysis.averageTStatistic.

A histogram is plotted for the analysis in x. If x includes the analyses for several gene sets, one histogram is plotted for each of the gene sets.

The statistic value of the input set is depicted as a vertical line.

The most common graphical parameters can be supplied as vectors (one entry per analyzed gene set) to vary them between the different analyses. These parameters are: main, xlab, ylab.

Value

Returns a list with all the underlying data for the plotted histograms as invisible object.

See Also

geneSetAnalysis, predefinedAnalyses, gsAnalysis, evaluateGeneSetUncertainty, plot.uncertaintyResult

Examples

# load data
data(exampleData)
  res <- geneSetAnalysis(
  	# global parameters
  	dat = countdata,
  	geneSets = pathways[3], 
  	analysis = analysis.averageCorrelation(), 
  	# additional parameters for analysis.averageCorrelation
  	labs = labels,
  	p = 1,
  	numSamples = 10)
  
  # plot the histogram for the cell cycle control gene set
  hist(res, main = names(pathways[3]))

Example data to label samples in countdata (rows) in two classes 0 and 1

Description

Randomly generated example data to give samples in countdata a class label 0 or 1

Usage

data(exampleData)

Details

The data consists of a variable labels with a vector of 96 values. The values are a random sequence of 0's and 1's.

Examples

data(exampleData)

print(labels)

Merge multiple probes for one gene

Description

Merges all probes belonging to the same gene by identifying duplicate row names in a data matrix.

Usage

mergeProbesForGenes(dat,
	method = c("mean", "max", "min", "median"))

Arguments

Value

A matrix of the same structure as dat, but possibly with fewer rows if probes were merged.

See Also

geneSetAnalysis

Examples

dat <- matrix(1:6, nrow=3, ncol=2)
rownames(dat) <- c("g1", "g2", "g1")

newDat <- mergeProbesForGenes(dat, method = "mean")

Venn Euler Diagramm

Description

Parses a GMT file as downloadable from MSigDB (presented in Subramanian et al.) and returns a list of gene sets.

Usage

	parseGmt(file)

Arguments

Details

Parses a GMT file and returns a list of gene sets. Each list element named according to the included gene set. The gene set files can be downloaded from http://www.broadinstitute.org/gsea/msigdb.

Value

A named list of gene sets.

References

Subramanian, A., Tamayo, P., Mootha, V. K., Mukherjee, S., Ebert, B. L., Gillette, M. A., Paulovich, A., Pomeroy, S. L., Golub, T. R., Lander, E. S., Mesirov, J. P. (2005) Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Science of the United States of America, 102, 15545–15550.

See Also

geneSetAnalysis, predefinedAnalyses, gsAnalysis

Example data for pathways to to gene set enrichtment analyses

Description

A list comprising exemplary pathways. Each pathway is represented by a vectro of random genes from countdata which are associated to this pathway.

Usage

data(exampleData)

Details

The data consists of a variable pathways of class list with 9 exemplary pathways. Each element of the list corresponds to one pathway - represented by a vector of gene names which are involved in this pathway.

Examples

data(exampleData)

print(pathways)

Plots the results of an uncertainty analysis.

Description

For each percentage of original gene set genes, the quantiles of the distribution obtained by a resampling simulation are plotted. Significance threshold (quantile of the Null distribution) and the test statistic of the original gene set are drawn as horizontal lines.

Usage

## S3 method for class 'uncertaintyResult'
plot(x,
	signLevel = x$signLevel,
	addLegend = TRUE,
	addMinimalStability = FALSE,
	...)

Arguments

Details

The function plots the quantiles of the resampling distributions for evaluated degrees of fuzziness. It requires the significance assessment step of the enrichment analysis configuration (parameter significance or gsAnalysis) to be a computer-intensive testing procedure that yields a distribution of gene set statistic values under the null hypothesis. Predefined configurations for which this plot works are analysis.gsea, analysis.averageCorrelation and analysis.averageTStatistic.

Three lines, corresponding to the different qunatiles with one dot per fuzziness evaluation (k) are plotted for the analysis in x. The significance threshold is shown as a green horizontal line. The statistic value of the original input set is depicted as a red horizontal line.

If addMinimalStability is TRUE, the lower bound of the stability is ploted as a dotted line.

Value

No return value, called for creating a plot.

See Also

geneSetAnalysis, predefinedAnalyses, gsAnalysis, evaluateGeneSetUncertainty

Examples

# load data
data(exampleData)

  res <- evaluateGeneSetUncertainty(
  	# parameters for evaluateGeneSetUncertainty
  	dat = countdata,
  	geneSet = pathways[[1]],
  	analysis = analysis.averageCorrelation(),
  	numSamplesUncertainty = 10,
  	N = seq(0.1,0.9, by=0.1),
  	# additional parameters for analysis.averageCorrelation
  	labs = labels,
  	numSamples = 10)
  
  # plot the results for the cell cycle control gene set
  plot(res, addMinimalStability = TRUE)

Plot overlap of gene sets and core set

Description

Plots a Venn diagramm of the overlaps of the core set and gene sets in an overrepresentation analysis.

Usage

plotOverrepresentation(
	object,
	signLevel = object$signLevel,
	subset = NULL,
	aggregate = FALSE,
	ask = FALSE,
	...)

Arguments

Value

No return value, called for creating a plot.

See Also

geneSetAnalysis, predefinedAnalyses, gsAnalysis

Examples

	data(exampleData)
  # use the absolute correlation as a gene-level statistic
  stat <- abs(apply(countdata,1,cor,y = labels))
  # define the core set as the 25% genes with the highest correlation
  coreSet <- rownames(countdata)[tail(order(stat), 25)]
  
  # perform an overrepresentation analysis
  resOverrep <- geneSetAnalysis(
  	dat = countdata,
  	geneSets = pathways,
  	analysis = analysis.customOverrepresentation(),
  	coreSet = coreSet,
  	adjustmentMethod = "fdr")
  
  # plot a Venn diagram
  plotOverrepresentation(resOverrep, subset = 1:3, aggregate = TRUE)

Predefined enrichment analyses

Description

Predefined analysis configurations that can be used in geneSetAnalysis

Usage

analysis.gsea()
analysis.overrepresentation()
analysis.customOverrepresentation()
analysis.averageCorrelation()
analysis.averageTStatistic()
analysis.globalTest()
analysis.globalAncova()

Details

The above functions return configurations for state-of-the-art analysis pipelines that can be used in geneSetAnalysis. All configurations are preconfigured collections of standard methods for the different pipeline steps. The following lists the methods chosen for the different steps and their parameters. For more detailed descriptions of these methods, please refer to the linked manual pages.

• analysis.gsea defines the Gene Set Enrichment Analysis (GSEA) method by Subramanain et al. Here, the gene-level statistic the absolute correlation calculated by gls.cor with the associated parameters labs, method and a preprocessing by transformation.abs. As a gene set statistic, the enrichment score (function gss.enrichmentScore with parameter p) is calculated. The significance is assessed in a permutation test using significance.permutation with testAlternative = "greater" and free parameter numSamples, labs.

• analysis.overrepresentation calculates an overrepresentation analysis using the gene-level statistic gls.tStatistic with parameters pValue(should be TRUE), alternative and labs. The resulting values are then transformed via transformation.adjustAndBinarize (parameters are the adjMethod and threshold). Finally gss.fisherExactTest is used as gene set statistic.

• analysis.customOverrepresentation calculates an overrepresentation analysis using a user-defined core set coreSet. That is, instead of calculating this core set internally based on differential expression as the standard overrepresentation analysis, this function allows for defining custom core sets. It internally uses the global analysis global.overrepresentation.

• analysis.averageCorrelation calculates the gene-level statistic as the absolute correlation using gls.cor (with parameters labs, method) and transformation.abs. The gene set statistic is the mean correlation calculated by gss.mean. The significance is assessed by comparing the gene set statistic to randomly sampled gene sets using significance.sampling (with the parameter numSamples and the preset parameter testAlternative = "greater").

• analysis.averageTStatistic uses the absolute t statistic as the gene-level statistic by applying gls.tStatistic (with parameters labs, pValue, alternative) and transformation.abs. The gene set statistic is the mean t statistic in the gene set as returned by gss.mean. The significance is assessed by comparing the gene set statistic to randomly sampled gene sets using significance.sampling (with the parameter numSamples and the preset parameter testAlternative = "greater").

• analysis.globalTest performs a global gene set enrichment analysis by Goeman et al. by applying the global.test function which in turn wraps the gt function in the globaltest package.

• analysis.globalAncova applies the global ANCOVA method by Hummel et al. using the global method global.ancova which wraps the GlobalAncova function in the GlobalAncova package.

Value

All functions return an object of class gsAnalysis that specifies the corresponding analysis parameters for geneSetAnalysis.

References

Subramanian, A., Tamayo, P., Mootha, V. K., Mukherjee, S., Ebert, B. L., Gillette, M. A., Paulovich, A., Pomeroy, S. L., Golub, T. R., Lander, E. S., Mesirov, J. P. (2005) Gene set enrichment analysis: a knowledge-based approach for interpreting genome-wide expression profiles. Proceedings of the National Academy of Science of the United States of America, 102, 15545–15550.

Hummel, M., Meister, R., Mansmann, U. (2008) GlobalANCOVA: exploration and assessment of gene group effects. Bioinformatics, 24(1), 78–85.

Goeman, J. J., van de Geer, S. A., de Kort, F., van Houwelingen, H. C. (2004) A global test for groups of genes: testing association with a clinical outcome. Bioinformatics, 20(1), 93–99.

See Also

geneSetAnalysis, gsAnalysis

Examples

	data(exampleData)
  # apply a gene set analysis based on the average absolute correlation
  resAvCor <- geneSetAnalysis(
  	# parameters for geneSetAnalysis
  	dat = countdata,
  	geneSets = pathways[1],
  	analysis = analysis.averageCorrelation(),
  	adjustmentMethod = "fdr",
  	# additional parameters for analysis.averageCorrelation
    labs = labels,
  	method = "pearson",
  	numSamples = 10)
  
  # apply an overrepresentation analysis
  resOverrep <- geneSetAnalysis(
  	# parameters for geneSetAnalysis
  	dat = countdata,
  	geneSets = pathways,
  	analysis = analysis.overrepresentation(),
  	adjustmentMethod = "fdr",
  	# additional parameters for analysis.overrepresentation
  	pValue = TRUE,
  	threshold = 0.1,
  	labs = labels
  )
  
  # apply a global analysis using GlobalAncova
  resGA <- geneSetAnalysis(
  	# parameters for geneSetAnalysis
  	dat = countdata,
  	geneSets = pathways[1],
  	analysis = analysis.globalAncova(),
  	adjustmentMethod = "fdr",
  	# additional parameters for analysis.globalAncova
  	labs = labels,
  	method = "approx")

Eliminate unknown genes from gene sets

Description

This function removes all genes that are not part of the experiment (not in rownames(dat)) from the specified gene sets which. All names are set to lower case.

Usage

preprocessGs(
	dat,
	geneSets)

Arguments

Value

A list of preprocessed gene sets, where each gene set only contains those genes that are also present in dat

See Also

geneSetAnalysis

Examples

#values are not important, only the row names are used
dat <- matrix(0,100,10)
rownames(dat) <- paste("gene",1:100,sep="")

geneSets <- list(
		gs1 = paste("GENE",1:20,sep=""),# all genes in the analyzed data
		gs2 = paste("Gene",101:110,sep=""),#no gene in the analyzed data
		gs3 = paste("gene",90:110,sep="")#some genes in the analyzed data
	)

newGeneSets <- preprocessGs(dat = dat, geneSets = geneSets)

Summarize gene set analysis results

Description

Prints a summary of a gene set analysis result object.

Usage

## S3 method for class 'gsaResult'
summary(object,
		mode = c("summary", "table"),
		orderBy = c("adjustedPValues", "rawPValues", "geneSetName"),
		significantOnly = FALSE,
		signLevel = object$signLevel,
		...)

Arguments

Value

Returns a data.frame with a summary of the produced results such as data set name, p-values, size of the given gene set.

See Also

geneSetAnalysis, hist.gsaResult, createSummaryTable

Examples

data(exampleData)
  # perform gene set analyses for several pathways
  res <- geneSetAnalysis(
  	# global parameters
  	dat = countdata,
  	geneSets = pathways, 
  	analysis = analysis.averageCorrelation(), 
  	# additional parameters for analysis.averageCorrelation
  	labs = labels,
  	numSamples = 10)
  
  #summarize the analyses
  summary(res, mode = "summary")
  
  summary(res, mode = "table", orderBy = "rawPValues")