[BioC] averaging duplicate spots after vsn transformation
huber at ebi.ac.uk
Sun Mar 16 15:58:28 CET 2008
yes, vsn is intended to be applied to the raw spot intensities values,
and replicate spots can be averaged afterwards.
vsn itself would probably work more or less the same either way, but:
one of the properties of raw microarray data is that the distribution is
often skewed (long upper tail), and one of the effects of (g)log
transformation is to make the data distribution more symmetric -- which
is a good thing for subsequent averaging, the resulting value will be a
better estimate of the "location" of the distribution than if you
averaged before transformation.
Wolfgang Huber EBI/EMBL Cambridge UK http://www.ebi.ac.uk/huber
Maurice Melancon wrote 13/03/2008 01:14:
> Hello all,
> I'm using vsn to transform a single-colour microarray experiment with 4 lab
> lines collected a 5 times, and three biological replicates (60 slides).
> Each EST is printed in duplicate, sometimes adjacent to one another and at
> other times at the ends of rows and beginning of the following row.
> Is it valid to vsn-transform the dataset prior to averaging the duplicate
> spots? The few papers I have that explicitly address minute details of data
> preparation do not present consensus - some people average before
> normalization, and others use normalization to decrease the variance among
> duplicate spots.
> It seems to me that providing as much information as possible would be of
> benefit to stabilizing variance in a dataset, but on the other hand it seems
> that pseudoreplication might bias vsn's accounting for differetially
> expressed features. I'm not much of a statistician, and we're an isolated
> facility so I hope that someone can shed light on what is becoming a lively
> With thanks,
> [[alternative HTML version deleted]]
> Bioconductor mailing list
> Bioconductor at stat.math.ethz.ch
> Search the archives: http://news.gmane.org/gmane.science.biology.informatics.conductor
More information about the Bioconductor