expressopdnn: Position Dependant Nearest Neighbors model for affy

Description

A wrapper to perform the PDNN method.

Usage

pdnn.scalevalue.exprSet(eset, scale.to=500)
expressopdnn(abatch, # background correction bg.correct = FALSE, bgcorrect.method = NULL, bgcorrect.param = list(), # normalize normalize = FALSE, normalize.method = NULL, normalize.param = list(),
              pmcorrect.method = c("pdnn", "pdnnpredict"),
      # pdnn findparams.param = list(),           # expression values summary.subset = NULL, # PDNN expression values scaling eset.normalize = TRUE, scale.to = 500, # misc. verbose = TRUE)

Arguments

abatch

object of AffyBatch-class.

bg.correct

a boolean to express whether background correction is wanted or not.

bgcorrect.method

the name of the background adjustment method.

bgcorrect.param

a list of parameters for bgcorrect.method (if needed/wanted).

eset

an object of ExpressionSet-class.

normalize

normalization step wished or not.

normalize.method

the normalization method to use.

normalize.param

a list of parameters to be passed to the normalization method (if wanted).

pmcorrect.method

the name of the PM adjustement method (only two choices here, default to 'pdnn').

findparams.param

a list of parameters to be passed to find.params.pdnn.

eset.normalize

is any normalization step on expression values to be performed.

scale.to

a value to scale against.

summary.subset

a list of 'affyids'. If NULL, then an expression summary value is computed for everything on the chip.

verbose

logical value. If TRUE it writes out some messages.

Value

An object of ExpressionSet-class, with an attribute pps.warnings as returned by the method computeExprSet.

Details

expressopdnn is very similar to expresso. It is mainly a wrapper around the pre-processing steps `background correction', `normalization', `perfect match correction' and the PDNN method to compute expression values (see the first reference for more details about the preprocessing steps and and the second reference for further details about the PDNN method).

The wrapper expresso has no way to handle easily the computation of chip-wide results that have to be used during the computeExprSet step. An easy way to overcome this was to write this simple wrapper.

pdnn.scalevalue is performed after the expression values have computed to somehow `normalize' the values between different chips. When setting normalize to TRUE this step might be considered unnecessary (and the eset.normalize set to FALSE).

Examples

Run this code

## load pre-computed parameters
data(hgu95av2.pdnn.params)

library(affydata)
data(Dilution)

## one CEL to go faster
afbatch <- Dilution[, 1]

## Take only few IDs (the 10 first)
ids <- ls(getCdfInfo(afbatch))[1:10]
eset <- expressopdnn(afbatch, bg.correct=FALSE,
                     normalize=FALSE,
                     findparams.param=list(params.chiptype=hgu95av2.pdnn.params,
                                           give.warnings=FALSE),
                     summary.subset=ids)