[BioC] LIMMA No residual using rmaPLM

[Ben Bolstad]

>
> Hello
>
> I have a general question regarding pre-processing data using RMA.
>
> Due to memory constraints, I am using the function justRMA() instead of
> fitPLM() for pre-processing my data (I have a couple of hundred samples so
> far). Is there a way to obtain standard errors of expression intensity
> estimates using this function? In this case, se.exprs returns a matrix of
> NA's.

Unfortunately neither justRMA() nor rma() return SE values.

> Also, I understand that these functions do not readily provide p-values
> associated with expression intensities (similar to the ones from MAS5 for
> example). I am wondering what would be a good way to check data quality in
> this case. Do you have any suggestions?


The P-values that MAS 5 produces are in relation to their presence absence
calls (not the expression value itself). There is no similar analog for
RMA in this sense. Generally speaking the MAS5 presence/absence calls are
pretty decent (although the MAS5 expression values have their faults) and
you could use these with RMA expression values if you wished without much
problem. That said, the reason people typically use these is to screen out
"noisy" genes at the low absolute intensity level. This tends to not be a
problem with RMA expression values.


You could do preliminary quality examination of things like chip pseudo
images using recent versions of RMAExpress found at:

http://www.stat.berkeley.edu/~bolstad/RMAExpress/RMAExpress.html

The latest alpha versions support extremely large datasets on memory
limited systems. Unfortunately things like RLE and NUSE (which are in
affyPLM) are not currently implemented in RMAExpress.