Merge two ExpressionSet objects by doing quantile normalization and computing partial residuals (i.e. substracting group mean expression in each batch). As currently implemented the method is only valid for balanced designs, e.g. each batch has the same number of samples per group.

Description

mergeBatches combines x and y into an ExpressionSet, performs quantile normalization and adjusts for batch effects by subtracting the mean expression in each batch (and then adding the grand mean so that the mean expression per gene is unaltered).

Usage

mergeBatches(x, y, mc.cores=1)

Arguments

Value

When y is an ExpressionSet, mergeBatches returns an ExpressionSet with combined expressions. Its featureData contains a variable "batch" indicating the batch that each sample corresponded to.

When y is a simulatedSamples object, mergeBatches is applied to combine x with each dataset in y and a list of ExpressionSet objects is returned.

Author(s)

David Rossell

Examples

#Fake data from 2 batches
x <- matrix(rnorm(6),nrow=2)
colnames(x) <- paste('x',1:3,sep='')
y <- matrix(1+rnorm(6),nrow=2)
colnames(y) <- paste('y',1:3,sep='')
x <- new("ExpressionSet",exprs=x)
y <- new("ExpressionSet",exprs=y)
exprs(x)
exprs(y)

#Merge & adjust
z <- mergeBatches(x,y)
exprs(z)