grooMethy: Groom methylation data to fix potential data issues
In REMP: Repetitive Element Methylation Prediction

Description Usage Arguments Details Value Examples

grooMethy is used to automatically detect and fix data issues including zero beta value, missing value, and infinite value.

grooMethy(
  methyDat,
  Seq.GR = NULL,
  impute = TRUE,
  imputebyrow = TRUE,
  mapGenome = FALSE,
  verbose = FALSE
)

`methyDat`	A `RatioSet`, `GenomicRatioSet`, `DataFrame`, `data.table`, `data.frame`, or `matrix` of Illumina BeadChip methylation data (450k or EPIC array) or Illumina methylation percentage estimates by sequencing. If the data are prepared as a `data.frame` or alike format, for Illumina array data, please make sure there is a column or row names are available to indicate the Illumina probe names (i.e. cg00000029); for sequencing methylation data, please provide the corresponding CpG location information in `Seq.GR`.
`Seq.GR`	A `GRanges` object containing genomic locations of the CpGs profiled by sequencing platforms. This parameter should not be `NULL` if the input methylation data `methyDat` are obtained by sequencing platform. The order of `Seq.GR` should match the order of `methyDat`. Note that the genomic location can be in either hg19 or hg38 build. See details.
`impute`	If `TRUE`, K-Nearest Neighbouring imputation will be applied to fill the missing values. Default = `TRUE`. See Details.
`imputebyrow`	If `TRUE`, missing values will be imputed using similar values in row (i.e., across samples); if `FALSE`, missing values will be imputed using similar values in column (i.e., across CpGs). Default is `TRUE`.
`mapGenome`	Logical parameter. If `TRUE`, function will return a `GenomicRatioSet` object instead of a `RatioSet`. This function is not applicable for sequencing data.
`verbose`	Logical parameter. Should the function be verbose?

For methylation data in beta value, if zero/one value exists, the logit transformation from beta to M value will produce infinite value. Therefore, zero/one beta value will be replaced with the smallest non-zero beta/largest non-one beta value found in the dataset. grooMethy can also handle missing value (i.e. NA or NaN) using KNN imputation (see impute.knn). The infinite value will be also treated as missing value for imputation. If the original dataset is in beta value, grooMethy will first transform it to M value before imputation is carried out. If the imputed value is out of the original range (which is possible when imputebyrow = FALSE), mean value will be used instead. Warning: imputed values for multimodal distributed CpGs (across samples) may not be correct. Please check package ENmix to identify the CpGs with multimodal distribution. Please note that grooMethy is also embedded in remp so the user can run remp directly without explicitly running grooMethy. For sequencing methylation data, please specify the genomic location of CpGs in a GenomicRanges object and specify it in Seq.GR. For an example of Seq.GR, Please run minfi::getLocations(IlluminaHumanMethylation450kanno.ilmn12.hg19) (the row names of the CpGs in Seq.GR can be NULL). The user should make sure the genome build of Seq.GR match the build specified in genome parameter of function initREMP and remprofile (default is "hg19").

A RatioSet or GenomicRatioSet containing beta value and M value of the methylation data.

# Get GM12878 methylation data (450k array)
if (!exists("GM12878_450k")) GM12878_450k <- getGM12878("450k")
GM12878_450k <- grooMethy(GM12878_450k, verbose = TRUE)

# Also works if data input is a matrix
grooMethy(minfi::getBeta(GM12878_450k), verbose = TRUE)