[BioC] Lumi normalization and VST help

[Paul Leo]

HI Christina,
I don't bother with beadstudio background subtraction, and have in found
that background subtraction in beadstudio or other background
subtraction algorithms on illumine chips make little or no difference
(in the experiments I have tested it on). But it won't hurt if you do
it.

Dump unnormalised data from beadstudio, most use the probe profile and
not the gene profile. Proceed as you have described. I don't think the
main issue is with VST and gene/probe profile; it is if you want to mix
together potential splice variants together and call them a "gene"  (or
perhaps mix crappy probes and good probes together and call the average
better). This is why most recommend using the probe profile but the
results will probably be very similar either way.

Yes you absolutely have to do normalization : treat each of the 12
separate arrays on the chip as being different (which they truly are).
Normalization is needed to correct for more than just technical scanning
artifacts that *may* be reduced if the arrays are on the same piece of
silicon. You will find PLENTY of variation in the arrays on the same
chip that normalization will be needed for!

Sound like you're on the right track follow the lumi pdf with the
default parameters and you should get something quite reasonable.

Cheers
Paul


-----Original Message-----
From: bioconductor-bounces at stat.math.ethz.ch
[mailto:bioconductor-bounces at stat.math.ethz.ch] On Behalf Of Christina
Markunas
Sent: Wednesday, April 01, 2009 11:50 AM
To: bioconductor at stat.math.ethz.ch
Subject: [BioC] Lumi normalization and VST help

Hi- 

 

I am new to analyzing Illumina microarray data.  We ran 6 case and 6
control samples (no replicates) on 1 Illumina HT-12 v3 Expression
BeadChip.  I would like to analyze our data using lumi.  I was planning
on doing the following: 1) background subtraction in Beadstudio, 2) VST
in lumi, and 3) quantile normalization in lumi.  I noticed that the
normalization methods in lumi are for between chip normalization.  Would
it still be appropriate to use quantile normalization in my situation
since I only have one chip (ie: between array normalization)?   

 

Also, will the VST method yield similar results if gene-level data are
used instead of probe-level data?

 

Thanks for your time,

 

Christina Markunas

cam27 at duke.edu



	[[alternative HTML version deleted]]

_______________________________________________
Bioconductor mailing list
Bioconductor at stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/bioconductor
Search the archives:
http://news.gmane.org/gmane.science.biology.informatics.conductor