|| at Feb 26, 2008 at 4:27 pm
You can find your answers from the lumi vignette (latest version is 1.5.17).
For your first question, you can use detectionCall function to select the
threshold of deteciton p-value cut-off (0.01 by default).
presentCount <- detectionCall(example.lumi, Th=0.01)
selDataMatrix <- dataMatrix[presentCount > 0,]
See the use case in the vignette for more details.
For the inverse transformation of VST, there is a function inverseVST. Now
it only works for the RSN or SSN normalization algorithms. Please see the
vignette for more details.
As for batch effects, it is always a problem for microarray analysis. That's
why we strongly suggest doing the experiment in the same batch. Different
normalization methods or modeling can help to reduce the batch effects
somehow. But they are always based on some assumptions. These assumptions
may not work for some cases.
On 2/26/08 5:00 AM, "bioconductor-request at stat.math.ethz.ch"
Date: Tue, 26 Feb 2008 00:26:22 +0100
From: "Simone de Jong" <email@example.com>
Subject: [BioC] Lumi: filtering and batch effects (combat)
Message-ID: <002a01c87805$d64fc220$56a32f0a at cog063>
Content-Type: text/plain; charset="us-ascii"
Since your responses helped me many times before, I'm posting two more
questions about Lumi:
* Can I filter genes that are below a certain detection pvalue out of my
file to speed up following analyses? How can I do that, at what stage and
what should be the threshold?
* I'm suspecting a batch effect in my data. I can work with ComBat, but I've
no idea how to get the adjusted values back into Lumi (a lumibatch) to go on
with transformation and normalization procedures. Any ideas?
Simone de Jong, Msc.