preprocess: Preprocess Raw ChIP-chip Intensities
In Ringo: R Investigation of ChIP-chip Oligoarrays

Description Usage Arguments Details Value Note Author(s) See Also Examples

Calls one of various (limma) functions to transform raw probe intensities into (background-corrected) normalized log ratios (M-values).

1 2	preprocess(myRG, method="vsn", ChIPChannel="R", inputChannel="G", returnMAList=FALSE, idColumn="PROBE_ID", verbose=TRUE, ...)

`myRG`	object of class `RGList`
`method`	string; denoting which normalization method to choose, see below for details
`ChIPChannel`	string; which element of the `RGList` holds the ChIP result, see details
`inputChannel`	string; which element of the `RGList` holds the untreated input sample; see details
`returnMAList`	logical; should an MAList object be returned? Default is to return an ExpressionSet object.
`idColumn`	string; indicating which column of the `genes` data.frame of the RGList holds the identifier for reporters on the microarray. This column, after calling `make.names` on it, will make up the unique `featureNames` of the resulting `ExpressionSet`. If argument `returnMAList` is `TRUE`, this argument is ignored.
`verbose`	logical; progress output to STDOUT?
`...`	further arguments to be passed on `normalizeWithinArrays` and `normalizeBetweenArrays`

The procedure and called limma functions depend on the choice of method.

loess: Calls normalizeWithinArrays with method="loess".
vsn: Calls normalizeBetweenArrays with method="vsn".
Gquantile: Calls normalizeBetweenArrays with method="Gquantile".
Rquantile: Calls normalizeBetweenArrays with method="Rquantile".
median: Calls normalizeWithinArrays with method="median".
nimblegen: Scaling procedure used by Nimblegen. Yields scaled log-ratios by a two step procedure: srat = log2(R) - log2(G) srat = srat - tukey.biweight(srat)
Gvsn: Learns vsn model on green channel intensities only and applies that transformation to both channels before computing fold changes.
Rvsn: Learns vsn model on red channel intensities only and applies that transformation to both channels before computing fold changes.
none: No normalization of probe intensities, takes raw log2(R)-log2(G) as component M and (log2(R)+log2(G))/2 as component A; uses normalizeWithinArrays with method="none".

Mostly with two-color ChIP-chip, the ChIP sample is marked with the red Cy5 dye and for the untreated input sample the green Cy3 dye is used. In that case the RGListmyRG's element R holds the ChIP data, and element G holds the input data. If this is not the case with your data, use the arguments ChIPChannel and inputChannel to specify the respective elements of myRG.

Returns normalized, transformed values as an object of class ExpressionList or MAList.

Since Ringo version 1.5.6, this function does not call limma's function backgroundCorrect directly any longer. If wanted by the user, background correction should be indicated as additional arguments passed on to normalizeWithinArrays or normalizeBetweenArrays, or alternatively call backgroundCorrect on the RGList before preprocessing.

Joern Toedling

normalizeWithinArrays, normalizeBetweenArrays, malist, ExpressionSet, vsnMatrix

   exDir <- system.file("exData",package="Ringo")
   exRG <- readNimblegen("example_targets.txt","spottypes.txt",
                         path=exDir)
   exampleX <- preprocess(exRG)
   sampleNames(exampleX) <- make.names(paste(exRG$targets$Cy5,"vs",
                                        exRG$targets$Cy3,sep="_"))
   print(exampleX)
   ### compare VSN to NimbleGen's tukey-biweight scaling
   exampleX.NG <- preprocess(exRG, method="nimblegen")
   sampleNames(exampleX.NG) <- sampleNames(exampleX)
   if (interactive())
     corPlot(cbind(exprs(exampleX),exprs(exampleX.NG)),
       grouping=c("VSN normalized","Tukey-biweight scaled"))