summarize: Tools for microarray preprocessing.

Description

These are tools to preprocess microarray data. They include background correction, normalization and summarization methods.

Usage

backgroundCorrectionMethods()
normalizationMethods()
summarizationMethods()
backgroundCorrect(object, method=backgroundCorrectionMethods(), copy=TRUE, extra, subset=NULL, target='core', verbose=TRUE)
summarize(object, probes=rownames(object), method="medianpolish", verbose=TRUE, ...)
## S3 method for class 'FeatureSet':
normalize(object, method=normalizationMethods(), copy=TRUE, subset=NULL,target='core', verbose=TRUE, ...)
## S3 method for class 'matrix':
normalize(object, method=normalizationMethods(), copy=TRUE, verbose=TRUE, ...)
## S3 method for class 'ff_matrix':
normalize(object, method=normalizationMethods(), copy=TRUE, verbose=TRUE, ...)
normalizeToTarget(object, targetDist, method="quantile", copy=TRUE, verbose=TRUE)

Arguments

object

Object containing probe intensities to be preprocessed.

method

String determining which method to use at that preprocessing step.

targetDist

Vector with the target distribution

probes

Character vector that identifies the name of the probes represented by the rows of object.

copy

Logical flag determining if data must be copied before processing (TRUE), or if data can be overwritten (FALSE).

subset

Not yet implemented.

target

One of the following values: 'core', 'full', 'extended', 'probeset'. Used only with Gene ST and Exon ST designs.

extra

Extra arguments to be passed to other methods.

verbose

Logical flag for verbosity.

...

Arguments to be passed to methods.

Value

backgroundCorrectionMethods and normalizationMethods will return a character vector with the methods implemented currently.
backgroundCorrect, normalize and normalizeToTarget will return a matrix with same dimensions as the input matrix. If they are applied to a FeatureSet object, the PM matrix will be used as input.
The summarize method will return a matrix with length(unique(probes)) rows and ncol(object) columns.

Details

Number of rows of object must match the length of probes.

Examples

Run this code

ns <- 100
nps <- 1000
np <- 10
intensities <- matrix(rnorm(ns*nps*np, 8000, 400), nc=ns)
ids <- rep(as.character(1:nps), each=np)
bgCorrected <- backgroundCorrect(intensities)
normalized <- normalize(bgCorrected)
summarizationMethods()
expression <- summarize(normalized, probes=ids)
intensities[1:20, 1:3]
expression[1:20, 1:3]
target <- rnorm(np*nps)
normalizedToTarget <- normalizeToTarget(intensities, target)

if (require(oligoData) & require(pd.hg18.60mer.expr)){
  ## Example of normalization with real data
  data(nimbleExpressionFS)
  boxplot(nimbleExpressionFS, main='Original')
  for (mtd in normalizationMethods()){
    message('Normalizing with ', mtd)
    res <- normalize(nimbleExpressionFS, method=mtd, verbose=FALSE)
    boxplot(res, main=mtd)
  }
}