Question

I’m trying to produce a heatmap of some gene expression data. Currently, the data I’d like to visualize is stored in an expression set data type, but the samples are not grouped correctly; that is, control samples and experimental samples are not grouped together. Here’s an abbreviated work flow:

output <- //analysis of raw data eset
someGenes <- eset [output$Table1, ]
heatmap(exprs(someGenes))

Essentially, I’d like to rearrange the samples in the eset before plotting. I’ve tried to treat the sample names as “columns” but haven’t been able to alter their position within the eset. I’m quite new to R, so any and all help would be greatly appreciated.

Here’s a link to the expression set data type for reference: http://rss.acs.unt.edu/Rdoc/library/Biobase/html/class.ExpressionSet.html

Many thanks.

Answer 1

For a reproducible example, we load Biobase and the sample data

library(Biobase)
data(sample.ExpressionSet)

then take a look at what we’ve got:

> sample.ExpressionSet
ExpressionSet (storageMode: lockedEnvironment)
assayData: 500 features, 26 samples 
  element names: exprs, se.exprs 
protocolData: none
phenoData
  sampleNames: A B ... Z (26 total)
  varLabels: sex type score
  varMetadata: labelDescription
featureData: none
experimentData: use 'experimentData(object)'
Annotation: hgu95av2 

We can subset an ExpressionSet like a matrix, so for instance using indices or sampleNames, e.g., by creating a column index cidx from a shuffle of the current sampleNames

cidx = sample(sampleNames(sample.ExpressionSet))

cidx is just a character vector. And then we re-order the columns with

> sample.ExpressionSet[, cidx]
ExpressionSet (storageMode: lockedEnvironment)
assayData: 500 features, 26 samples 
  element names: exprs, se.exprs 
protocolData: none
phenoData
  sampleNames: Q C ... B (26 total)
  varLabels: sex type score
  varMetadata: labelDescription
featureData: none
experimentData: use 'experimentData(object)'
Annotation: hgu95av2 

So you’ve re-ordered the columns before plotting. But heatmap is, by default, going to cluster the samples based on their ‘distance’ between one another. So it doesn’t matter that you’ve re-ordered the samples, heatmap will figure out the same distances, and plot the same heatmap. You might want image, or to provide the argument Colv (see ?heatmap) or use heatmap.2 in the gplots package, or… Hope that helps.