normalizeByReference {tilingArray} | R Documentation |
Adjust the hybridization intensities from an oligonucleotide microarray for probe-specific response effect by using one or several reference hybridizations.
normalizeByReference(x, reference, pm, background, refSig, nrStrata=10, cutoffQuantile=0.05, plotFileNames, verbose=FALSE)
x |
ExpressionSet containing the data to be normalized. |
reference |
ExpressionSet with the same number of features as x ,
containing the reference signal, on the raw scale (non-logarithmic).
This argument can be used to directly input the data from a set of
replicate DNA hybridizations.
Alternatively, the argument refSig can be specified. |
pm |
Indices specifying the perfect match features in
reference (see Details).
This can be either an integer vector with values between 1 and
nrow(exprs(reference)) or a logical vector. |
background |
Indices specifying a set of background features in
x (see Details).
This can be either an integer vector with values between 1 and
nrow(exprs(x)) or a logical vector. |
refSig |
A numeric vector of the same length as pm with
estimates of probe response effects, on a logarithm-like scale.
This argument can be specified alternatively to reference . |
nrStrata |
Integer (length 1), number of strata for the estimation of the background function. |
cutoffQuantile |
Numeric (length 1), the probes whose reference signal is below this quantile are thrown out. |
plotFileNames |
Character vector whose length is the same as the
number of arrays in x . Optional, if missing, no plots are
produced. |
verbose |
Logical of length 1, if TRUE , some messages about
progress are printed. |
The intensities in x
are adjusted according to the
reference values.
Typically, the reference values are obtained by hybridizing a DNA
sample to the array, so that the abundance of target is
the same for all reference probes, and their signal can be used to estimate
the probe sequence effect. A reference probe is a probe that
perfectly matches the target genome exactly once.
Usually, not all probes on a chip are reference
probes, hence the subset of those that are is specified by the
argument pm
.
The background signal is estimated from the probes indicated by the
argument background
. They need to be a strict subset of the
reference
probes. I.e., they need to uniquely
match the target organism's DNA, but are not expected to match any of its transcripts. A
robust estimation method is used, so a small fraction of
background
probes that do hit transcripts is not harmful.
A limitation of this normalization method is that it only makes sense for the data from reference probes, NA values are returned for all other probes.
The functions PMindex
and BGindex
can be used to produce
the pm
and background
arguments from a probeAnno
environment such as provided in the davidTiling package.
To summarize, a reference probe (indicated by argument pm
) is a
probe that perfectly matches the target genome exactly once, a
background probe (indicated by argument background
)
is a reference probe which we expect not to be transcribed. These
should not be confused with what is called 'perfect match' and
'mismatch' probes in Affymetrix annotation.
A copy of x
with the normalized intensities.
W. Huber huber@ebi.ac.uk
Huber W, Toedling J, Steinmetz, L. Transcript mapping with high-density oligonucleotide tiling arrays. Bioinformatics 22, 1963-1970 (2006).
PMindex
, BGindex
## see vignette assessNorm.Rnw in inst/scripts directory