preprocessQuantile: Stratified quantile normalization for an Illumina methylation...
In minfi: Analyze Illumina Infinium DNA methylation arrays

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

Stratified quantile normalization for Illumina amethylation arrays.

This function implements stratified quantile normalization preprocessing for Illumina methylation microarrays. Probes are stratified by region (CpG island, shore, etc.)

preprocessQuantile(object, fixOutliers = TRUE, removeBadSamples = FALSE,
                   badSampleCutoff = 10.5, quantileNormalize = TRUE,
                   stratified = TRUE, mergeManifest = FALSE, sex = NULL,
                   verbose = TRUE)

`object`	An object of class `RGChannelSet` or `[Genomic]MethylSet`.
`fixOutliers`	Should low outlier Meth and Unmeth signals be fixed?
`removeBadSamples`	Should bad samples be removed?
`badSampleCutoff`	Samples with median Meth and Umneth signals below this cutoff will be labelled ‘bad’.
`quantileNormalize`	Should quantile normalization be performed?
`stratified`	Should quantile normalization be performed within genomic region strata (e.g. CpG island, shore, etc.)?
`mergeManifest`	Should the information in the associated manifest package be merged into the output object?
`sex`	Gender
`verbose`	Should the function be verbose?

This function implements stratified quantile normalization preprocessing for Illumina methylation microarrays. If removeBadSamples is TRUE we calculate the median Meth and median Unmeth signal for each sample, and remove those samples where their average falls below badSampleCutoff. The normalization procedure is applied to the Meth and Unmeth intensities separately. The distribution of type I and type II signals is forced to be the same by first quantile normalizing the type II probes across samples and then interpolating a reference distribution to which we normalize the type I probes. Since probe types and probe regions are confounded and we know that DNAm distributions vary across regions we stratify the probes by region before applying this interpolation. For the probes on the X and Y chromosomes we normalize males and females separately using the gender information provided in the sex argument. If gender is unspecified (NULL), a guess is made using by the getSex function using copy number information. Background correction is not used, but very small intensities close to zero are thresholded using the fixMethOutlier. Note that this algorithm relies on the assumptions necessary for quantile normalization to be applicable and thus is not recommended for cases where global changes are expected such as in cancer-normal comparisons.

Note that this normalization procedure is essentially similar to one previously presented (Touleimat and Tost, 2012), but has been independently re-implemented due to the present lack of a released, supported version.

a GenomicRatioSet

A bug in the function was found to affect the Beta values of type I probes, when stratified=TRUE (default). This is fixed in minfi version 1.19.7 and 1.18.4 and greater.

Rafael A. Irizarry

N Touleimat and J Tost. Complete pipeline for Infinium Human Methylation 450K BeadChip data processing using subset quantile normalization for accurate DNA methylation estimation. Epigenomics (2012) 4:325-341.

getSex, minfiQC, fixMethOutliers for functions used as part of preprocessQuantile.

if (require(minfiData)) {
  # NOTE: RGsetEx.sub is a small subset of RGsetEx; only used for computational
  #       speed
  GMset.sub.quantile <- preprocessQuantile(RGsetEx.sub)
}
## Not run: 
if(require(minfiData)) {
  GMset <- preprocessQuantile(RGsetEx)
}

## End(Not run)