gQTLstats (version 1.4.2)

clipPCs: transformations of expression data in smlSet instances

Description

transformations of expression data in smlSet instances or assay data in RangedSummarizedExperiment

Usage

clipPCs(x, inds2drop, center = TRUE)
regressOut(x, rhs, ...)

Arguments

x

a RangedSummarizedExperiment object

inds2drop

Vector of PCs to be eliminated by setting the associated diagonal elements in the SVD to zero before recomposing the matrix of expression values. If the value 0 is present in inds2drop, the smlSet is returned unchanged, with a message.

center

logical, passed to prcomp

rhs

formula fragment (no dependent variable) used to form residuals in a reexpression of the expression matrix; variable bindings found in pData of an ExpressionSet or colData of a RangedSummarizedExperiment

...

arguments passed to lmFit

Value

• a RangedSummarizedExperiment object

Details

clipPCs is an operation on the n x p transposed matrix X of expression data. The singular value decomposition X = UDV^t is formed, the diagonal elements of D corresponding to inds2drop are set to zero yielding the diagonal matrix E, and then Y = UEV^t is computed and transposed to replace the expression data.

regressOut obtains residuals after genewise regression of expression on the design matrix specified by the rhs; lmFit is used to compute coefficients, linear predictions and residuals.

References

The use of PCA-based adjustments to remove mass extraneous effects from expression matrices has been criticized in work of Oliver Stegle and Jeffrey Leek, who offer Bayesian PEER and SVA respectively as alternative solutions.

Examples

if(require(geuvPack)){
 data(geuFPKM)
 cg = clipPCs(geuFPKM, 1:10)
 ro = regressOut(cg, ~popcode)
 ro
 }