4 RMA
RMA is a popular model-based approach to normalization and calculating
gene expression for Affymetrix microarrays. The approach consists of
background correction, quantile normalization and a modeling
probe-specific effects across multiple arrays using a median-polish
method for fitting the model.
The options for this method include:
- Use Background Correction
This is almost always used, but can be disabled for experimentation.
- Normalization Method
There are two different methods of normalizing the probe-level PM
data: quantile normalization in which the intensities are replaced
with mean intensity at that quantile, and a mean normalization in
which mean PM intensity is set equal for all chips. The other option
is to use no normalization (not generally recommended). The default is
quantile normalization.
- Normalize AFFX Probesets
This allows the AFFX probesets to be ignored during quantile normalization,
which could potentially avoid a bias due to vastly different signal across
chips in these control probesets.
Finally, you can select the name of the output file to write
expression results. This file is a tab-delimited output file with
columns labeled with CEL file names (without the .CEL extension) and
rows corresponding to probesets. The default is exprs-rma.txt
in the working directory.