parameterizeRMA: Calculate parameters for incremental RMA

In steveneschrich/IncrementalRMA: Affymetrix-based RMA Processing Using Fixed Parameters

Description

Calculate and store parameters for future use in incrementalRMA.

Usage

parameterizeRMA(abatch)

Arguments

abatch

An AffyBatch-class to calculate parameters on.

Details

IncrementalRMA works by applying previous computed parameters from a full RMA process. This function calculates the necessary parameters for future incrementalRMA use.

RMA consists of three steps: background correction, quantile normalization and median polish. Background correction is done per-chip, so there is no need for pre-calculating parameters. Quantile normalization computes values for each quantile in expression data and then sets the intensity distribution equal to these quantiles. Median polish sweeps out row-wise and column-wise medians from probe expression; column-wise effects are used as expression estimates.

Given the above description, quantiles (from quantile normalization) and row-wise effects (from median polish) can be computed on a set of samples and stored. With this information, incrementalRMA can then calculate gene expression from a new sample without re-computing these parameters.

This function computes these parameters and stores them in a list, such that the list can then be used. The elements of the list currently include:

• probeEffectsRow-wise effects from median polish step.

• normalizationVectorQuantile values from quantile normalization step.

• referenceCELFilesAn AffyBatch-class containing the reference cel files (for error calculations).

A simple test of incrementalRMA is to parameterizeRMA, then apply incrementalRMA and compare with using rma on the set. See the example for code to test this.

Value

A list containing

Examples

## Not run: 
params <- parameterizeRMA(abatch)

all.equal(exprs(incrementalRMA(abatch, parameterizeRMA(abatch))), exprs(affy::rma(abatch)))

## End(Not run)

steveneschrich/IncrementalRMA documentation built on Dec. 23, 2021, 5:32 a.m.