RUVfit: Remove unwanted variation when testing for differential...
In ChuanJ/test: Analysing Illumina HumanMethylation BeadChip Data

Description Usage Arguments Details Value Author(s) References See Also Examples

Provides an interface similar to lmFit from limma to the RUV2, RUV4, RUVinv and RUVrinv functions from the ruv package, which facilitates the removal of unwanted variation in a differential methylation analysis. A set of negative control variables, as described in the references, must be specified.

1	RUVfit(Y, X, ctl, Z = 1, k = NULL, method=c("inv", "rinv", "ruv4", "ruv2"), ...)

`Y`	numeric `matrix` with rows corresponding to the features of interest such as CpG sites and columns corresponding to samples or arrays.
`X`	The factor(s) of interest. A m by p matrix, where m is the number of samples and p is the number of factors of interest. Very often p = 1. Factors and dataframes are also permissible, and converted to a matrix by `design.matrix`.
`ctl`	logical vector, `length == nrow(Y)`. Features that are to be used as negative control variables are indicated as TRUE, all other features are FALSE.
`Z`	Any additional covariates to include in the model, typically a m by q matrix. Factors and dataframes are also permissible, and converted to a matrix by `design.matrix`. Alternatively, may simply be 1 (the default) for an intercept term. May also be NULL.
`k`	integer, required if `method` is "ruv2" or "ruv4". Indicates the number of unwanted factors to use. Can be 0.
`method`	character string, indicates which `ruv` method should be used.
`...`	additional arguments that can be passed to `RUV2`, `RUV4`, `RUVinv` and `RUVrinv`. See linked function documentation for details.

This function depends on the ruv package and is used to estimate and adjust for unwanted variation in a differential methylation analysis. Briefly, the unwanted factors W are estimated using negative control variables. Y is then regressed on the variables X, Z, and W. For methylation data, the analysis is performed on the M-values, defined as the log base 2 ratio of the methylated signal to the unmethylated signal.

A list containing:

`betahat`	The estimated coefficients of the factor(s) of interest. A p by n matrix.
`sigma2`	Estimates of the features' variances. A vector of length n.
`t`	t statistics for the factor(s) of interest. A p by n matrix.
`p`	P-values for the factor(s) of interest. A p by n matrix.
`Fstats`	F statistics for testing all of the factors in X simultaneously..
`Fpvals`	P-values for testing all of the factors in X simultaneously.
`multiplier`	The constant by which `sigma2` must be multiplied in order to get an estimate of the variance of `betahat`.
`df`	The number of residual degrees of freedom.
`W`	The estimated unwanted factors.
`alpha`	The estimated coefficients of W.
`byx`	The coefficients in a regression of Y on X (after both Y and X have been "adjusted" for Z). Useful for projection plots.
`bwx`	The coefficients in a regression of W on X (after X has been "adjusted" for Z). Useful for projection plots.
`X`	`X`. Included for reference.
`k`	`k`. Included for reference.
`ctl`	`ctl`. Included for reference.
`Z`	`Z`. Included for reference.
`fullW0`	Can be used to speed up future calls of `RUVfit`.
`include.intercept`	`include.intercept`. Included for reference.
`method`	Character variable with value indicating which RUV method was used. Included for reference.

Jovana Maksimovic jovana.maksimovic@mcri.edu.au

Gagnon-Bartsch JA, Speed TP. (2012). Using control genes to correct for unwanted variation in microarray data. Biostatistics. 13(3), 539-52. Available at: http://biostatistics.oxfordjournals.org/content/13/3/539.full. Gagnon-Bartsch, Jacob, and Speed. 2013. Removing Unwanted Variation from High Dimensional Data with Negative Controls. Available at: http://statistics.berkeley.edu/tech-reports/820.

RUV2, RUV4, RUVinv, RUVrinv, topRUV

if(require(minfi) & require(minfiData) & require(limma)) {
# Get methylation data for a 2 group comparison
meth <- getMeth(MsetEx)
unmeth <- getUnmeth(MsetEx)
Mval <- log2((meth + 100)/(unmeth + 100))
group <- factor(pData(MsetEx)$Sample_Group)
design <- model.matrix(~group)
# Perform initial analysis to empirically identify negative control features 
# when not known a priori
lFit <- lmFit(Mval,design)
lFit2 <- eBayes(lFit)
lTop <- topTable(lFit2,coef=2,num=Inf)
# The negative control features should *not* be associated with factor of interest
# but *should* be affected by unwanted variation 
ctl <- rownames(Mval) %in% rownames(lTop[lTop$adj.P.Val > 0.5,])
# Perform RUV adjustment and fit
fit <- RUVfit(Y=Mval, X=group, ctl=ctl)
fit2 <- RUVadj(Y=Mval, fit=fit)
# Look at table of top results
top <- topRUV(fit2)
}