Hi,everyone.
I am now analyzing my microarray data using limma. I am not sure
when to use between-array normalization.
In the first case of chapter 11 in limma user's guide, it is
mentioned that between-array normalization is not routine and should
only be done when there is good evidence. I wonder what is a good
evidence and whether I should carry out between-array normalization
in my data. See the attachment for the boxplot.
In the attached boxplot,the blue, red, and green ones stand for
three different hybridization groups. There are 6 replications in
each group. I want to carry out between-array normalization in
the three groups respectively and then combine them together to fit
a linear model. Is this reasonable? In my opinion, we can always
carry out between-array normalization between the replications of
the same hybridization group (unless there is dye swap).
Thanks in advance for your reply.
Dejian Zhao
De-Jian,ZHAO wrote:
> Hi,everyone.
>
> I am now analyzing my microarray data using limma. I am not sure
> when to use between-array normalization.
>
> In the first case of chapter 11 in limma user's guide, it is
> mentioned that between-array normalization is not routine and should
> only be done when there is good evidence. I wonder what is a good
> evidence and whether I should carry out between-array normalization
> in my data. See the attachment for the boxplot.
>
> In the attached boxplot,the blue, red, and green ones stand for
> three different hybridization groups. There are 6 replications in
> each group. I want to carry out between-array normalization in
> the three groups respectively and then combine them together to fit
> a linear model. Is this reasonable? In my opinion, we can always
> carry out between-array normalization between the replications of
> the same hybridization group (unless there is dye swap).
The rule-of-thumb is that if you are working with two-color data and
using ratios only, then you probably do not need between-array
normalization. If you want to do single-channel analyses of two-color
data or are using single-channel data (affy, etc.), then you need to
use
between-array normalization.
The other rule-of-thumb with normalization is to do the least
reasonable
amount possible. So, despite the fact that one _can_ do between-array
normalization, one should not do it unless there is a good evidence
that
it is important, after looking at within-array normalization results.
Sean
Sean Davis Wrote:
>The rule-of-thumb is that if you are working with two-color data and
>using ratios only, then you probably do not need between-array
>normalization.
What is the rationale behind this claim? Is it because the common ref
channel serves as kind of internal calibration control already?
Seems that the current recommendation of normalization is on the less
agreesive side?
-Simon
Dear Simon,
On Wednesday 24 October 2007 02:32, Simon Lin wrote:
> Sean Davis Wrote:
> >The rule-of-thumb is that if you are working with two-color data
and
> >using ratios only, then you probably do not need between-array
> >normalization.
>
> What is the rationale behind this claim? Is it because the common
ref
> channel serves as kind of internal calibration control already?
>
I think (at least part of) the argument had to do with the variance-
bias
trade-off. And I think the NAR paper by Dudoit et al. on normalization
had
some comments regarding these issues (though that paper is a few years
old by
now). I think I also remember seening a pdf of a talk by K. Kerr and
colleagues where these issues might have been addressed (though I am
not
sure).
Anyway, I'm also interested in knowing if there are stronger arguments
and
more recent studies.
Best,
R.
> Seems that the current recommendation of normalization is on the
less
> agreesive side?
>
> -Simon
>
> _______________________________________________
> 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
--
Ram?n D?az-Uriarte
Statistical Computing Team
Centro Nacional de Investigaciones Oncol?gicas (CNIO)
(Spanish National Cancer Center)
Melchor Fern?ndez Almagro, 3
28029 Madrid (Spain)
Fax: +-34-91-224-6972
Phone: +-34-91-224-6900
http://ligarto.org/rdiaz
PGP KeyID: 0xE89B3462
(http://ligarto.org/rdiaz/0xE89B3462.asc)
**NOTA DE CONFIDENCIALIDAD** Este correo electr?nico, y
...{{dropped:3}}
