piano (version 1.12.0)

diffExp: Perform differential expression analysis

Description

Identifies differentially expressed genes by using the linear model approach of limma. Optionally produces a Venn diagram, heatmap, Polar plot and volcano plot.

Usage

diffExp(arrayData, contrasts, chromosomeMapping, fitMethod = "ls", adjustMethod = "fdr", significance = 0.001, plot = TRUE, heatmapCutoff = 1e-10, volcanoFC = 2, colors=c("red","green","blue","yellow", "orange","purple","tan","cyan","gray60","black"), save = FALSE, verbose = TRUE)

Arguments

arrayData

an object of class ArrayData.

contrasts

a character vector giving the contrasts to be tested for differential expression. Use extractFactors to get allowed contrasts.

chromosomeMapping

character string giving the name of the chromosome mapping file, or an object of class data.frame or similar containing the chromosome mapping. Required for the Polar plot if the ArrayData object lacks annotation information. See details below.

fitMethod

character string giving the fitting method used by lmFit. Can be either "ls" for least squares (default) or "robust" for robust regression.

adjustMethod

character string giving the method to use for adjustment of multiple testing. Can be "holm", "hochberg", "hommel", "bonferroni", "BH", "BY", "fdr" (default) or "none". See p.adjust for details.

significance

number giving the significance cutoff level for the Venn diagram and the horizontal line drawn in the volcano plot. Defaults to 0.001.

plot

should plots be produced? Set either to TRUE (default) or FALSE to control all plots, or to a character vector with any combination of "venn", "heatmap", "polarplot" and "volcano", to control the single plots (e.g. plot=c("venn","polarplot") or plot="heatmap").

heatmapCutoff

number giving the significance cutoff level for the heatmap. Defaults to 1e-10.

volcanoFC

number giving the x-coordinates of the vertical lines drawn in the volcano plot. Defaults to 2.

colors

character vector of colors to be used by the Venn diagram and Polar plot.

save

should the figures and p-values be saved? Defaults to FALSE.

verbose

verbose? Defaults to TRUE.

Value

A list with elements:

pValues

data.frame containing adjusted p-values (according to argument adjustMethod) for each contrast

foldChanges

data.frame containing log2 fold changes for each contrast

resTable

a list with an element for each contrast, each being a data.frame with full result information

vennMembers

list containing the gene members of each area of the Venn diagram (only returned when a Venn diagram is drawn)

Details

This function uses limma to calculate p-values measuring differential expression in the given contrasts. The uniqueFactors given by extractFactors can be used to define a contrast vector, where each element should be a character string on the form "uniqueFactorA - uniqueFactorB", note the space surrounding the -. (See the example below and for extractFactors.) If appropriate annotation is missing for the ArrayData object the user can suppply this as chromosomeMapping. This should be either a data.frame or a tab delimited text file and include the columns chromosome with the chromosome name and chromosome location containing the starting position of each gene. A - sign can be used to denote the antisense strand but this will be disregarded while plotting. The rownames should be probe IDs or, if using a text file, the first column with a column header should contain the probe IDs. Note that the fitMethod="robust" may need longer time to run. A Venn diagram can be drawn for up to five contrasts (diffExp() will use vennDiagram). The heatmap shows normalized expression values of the genes that pass the heatmapCutoff in at least one contrast. A volcano plot is produced for each contrast showing magnitude of change versus significance. The Polar plot sorts the genes according to chromosomal location, for each chromosome starting with unknown positions followed by increasing number in the chromosome location column. Genes which do not map to any chromosome are listed as U for unknown. The radial lines in the Polar plot are -log10 scaled p-values, so that a longer line means a smaller p-value. This gives an overview of the magnitude of differential expression for each contrast.

Typical usages are:

# Identify significantly changed genes in 'm1' and 'm2' compared to 'wt':
diffExp(arrayData, contrasts=c("m1 - wt", "m2 - wt"))

References

Smyth, G. K. (2005). Limma: linear models for microarray data. In: 'Bioinformatics and Computational Biology Solutions using R and Bioconductor'. R. Gentleman, V. Carey, S. Dudoit, R. Irizarry, W. Huber (eds), Springer, New York, pages 397--420.

See Also

piano, loadMAdata, extractFactors, polarPlot, runGSA, limma, venn, heatmap.2

Examples

  # Get path to example data and setup files:
  dataPath <- system.file("extdata", package="piano")

  # Load normalized data:
  myArrayData <- loadMAdata(datadir=dataPath, dataNorm="norm_data.txt.gz", platform="yeast2")

  # Perform differential expression analysis:
  pfc <- diffExp(myArrayData, contrasts=c("aerobic_Clim - anaerobic_Clim",
                                          "aerobic_Nlim - anaerobic_Nlim"))
                                          
  # Order the genes according to p-values, for aerobic_Clim vs anaerobic_Clim:
  o <- order(pfc$resTable$'aerobic_Clim - anaerobic_Clim'$P.Value)
  
  # Display statistics for the top 10 significant genes:
  pfc$resTable$'aerobic_Clim - anaerobic_Clim'[o[1:10],]