One of the ways to deal with the presence of multiple probes for the same gene in a microarray is to take the average of the expression value of the different probes. This would provide a single representative expression value for that gene.
The avereps
function in limma
can be used for this purpose. The Note
section for the avereps
function from the limma reference manual states the following (emphasis added by me):
This function should only be applied to normalized log-expression values, and not to raw unlogged expression values. It will generate an error message if applied to RGList or EListRaw objects.
Would it not be more accurate to take the mean of normalized base10-expression values (and then convert to log2) rather than take the mean of the normalized log2-expression values as suggested by the note?
I was asking more about the difference between using normalized non-log values vs normalized log values. In the original post, by base10, I meant non-log (sorry for the confusion about that).
To illustrate with the example that you have shown, say we have two matrices
e1
ande2
, the first of which is non-log and the latter which is in log2According to the note in the limma reference value, it expects
e2
as the input because it states:But wouldn't it make more sense to give
e1
as the input and thenlog2
it after the averaging?No, it doesn't make sense to do that. Raw microarray intensity values have a right skew, and since the mean is not robust to outliers you want to take logs first to minimize that.
Anyway, for Affy data the convention is to take logs using the probe values, normalize, then use
medianpolish
to get the probeset expression values (that's the RMA algorithm). So you should normally be starting with normalized, log base 2 values. You would have to do something inadvisable to have unlogged expression values in the first place (the sample.ExpressionSet data are, IIRC MAS5.0 summarized data, and nobody has used MAS5.0 since maybe 2002 when Rafa showed it was bunk).That makes sense. Thank you for the detailed reply!