R/AffymetrixCelFile.fitQuantileNormFcn.R

In HenrikBengtsson/aroma.affymetrix: Analysis of Large Affymetrix Microarray Data Sets

###########################################################################/**
# @set "class=AffymetrixCelFile"
# @RdocMethod fitQuantileNormFcn
#
# @title "Fits quantile normalization functions for the arrays in the data set"
#
# \description{
#  @get "title".
# }
#
# @synopsis
#
# \arguments{
#   \item{yTarget}{The target probe signals.}
#   \item{subset}{An optional @numeric @vector specifying the indices of the
#      subset of probes to be used to fit the normalization function.}
#   \item{spar, nknots}{Control parameters passed to
#      @see "stats::smooth.spline".}
#   \item{...}{Not used.}
#   \item{verbose}{See @see "R.utils::Verbose".}
# }
#
# \value{
#  Returns a normalization @function.
# }
#
# @author "HB"
#
# \seealso{
#   @seeclass
# }
#*/###########################################################################
setMethodS3("fitQuantileNormFcn", "AffymetrixCelFile", function(this, yTarget, subset=NULL, ..., spar=NULL, nknots=1024, verbose=FALSE) {
  # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  # Validate arguments
  # - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
  # Argument 'verbose':
  verbose <- Arguments$getVerbose(verbose)

  # Argument 'yTarget':
  if (!identical(attr(yTarget, "isSorted"), TRUE)) {
    # Sort target signals
    verbose && enter(verbose, "Sorting ", length(yTarget), " target signals")
    yTarget <- sort(yTarget)
    attr(yTarget, "isSorted") <- TRUE
    verbose && exit(verbose)
  }

  # Argument 'controlParams':
#  spar <- controlParams$spar
#  nknots <- controlParams$nknots


  verbose && enter(verbose, "Fitting (quantile) normalization function")

  # Read the probe intensities
  y <- getData(this, fields="intensities", verbose=less(verbose))

  # Sort signals
  verbose && enter(verbose, "Sorting probe signals")
  y <- sort(y)
  verbose && exit(verbose)

  # Fit normalization function
  verbose && enter(verbose, "Fitting smooth spline")
  ok <- !is.na(yTarget) & !is.na(y)
  sp <- smooth.spline(x=y[ok], y=yTarget[ok], spar=spar, nknots=nknots,
                                                          keep.data=FALSE)
  verbose && exit(verbose)

  # Create a minimal 'smooth.spline' object for prediction.
  # Note: You can not do predict(sp$fit, ...) because then there
  # will be a problem with Recall().  See r-devel on 2006-04-05.
  fit <- structure(list(fit=sp$fit), class=class(sp))

  env <- new.env(parent=baseenv())
  assign("fit", fit, envir=env)
  # Create transformation function
  fcn <- function(x, ...) {
    stats::predict(fit, x, ...)$y
  }
  environment(fcn) <- env

  verbose && exit(verbose)

  fcn
}, private=TRUE) # fitQuantileNormFcn()