Hi,
We have been comparing the effects of a mutation to two different
duplications of the wild type gene (ie g- v g-/dp1(g+) and g- v
g-/dp2(g+)) and then doing rma on either the pooled affy chips or the
two genotypes separately and then pooling them for analysis in dChip.
But which of the methods, separate or together is the statistically
the correct way of doing the rma analysis? The two give different
numbers of differentially expressed genes.
The second question, does rma in affy v1.4.21 have the same bug as
justRMA and gcRMA in affy v1.4.22?
thanks,
Simon
For your first question, it is almost always preferable to run rma on
all samples in a set together rather than in two batches. For the
second
question, the bug only concerns justRMA and justGCRMA, so rma and
gcrma
are not affected.
Best,
Jim
James W. MacDonald
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109
734-647-5623
>>> Simon Kidd <kidd@mail.rockefeller.edu> 04/14/04 11:35AM >>>
Hi,
We have been comparing the effects of a mutation to two different
duplications of the wild type gene (ie g- v g-/dp1(g+) and g- v
g-/dp2(g+)) and then doing rma on either the pooled affy chips or the
two genotypes separately and then pooling them for analysis in dChip.
But which of the methods, separate or together is the statistically
the correct way of doing the rma analysis? The two give different
numbers of differentially expressed genes.
The second question, does rma in affy v1.4.21 have the same bug as
justRMA and gcRMA in affy v1.4.22?
thanks,
Simon
_______________________________________________
Bioconductor mailing list
Bioconductor@stat.math.ethz.ch
https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor
At 8.50am -0400 15/4/04, James MacDonald wrote:
>For your first question, it is almost always preferable to run rma on
>all samples in a set together rather than in two batches. For the
second
>question, the bug only concerns justRMA and justGCRMA, so rma and
gcrma
>are not affected.
>
Thanks, however my collaborator thinks I did not phrase my question
very well and did not emphasis enough that the genetic background of
the strains could be very different, so this is what she thinks the
question should be:
>We have performed two experiments. One compares fly strain A with
>mutation M to fly strain A with a duplication for the wild type gene
>M. The second experiment compares fly strain B with the same
>mutation M with a different duplication for the wild type gene M.
>Aside from the particular mutation we are assaying the genetic
>background of the strains could be very different. If one
>normalizes arrays from different tissues using dChip, it is
>suggested that the arrays get normalized separately. Is this also
>true for RMA or should they be normalized together?
If the strains are very different should they still be normalised
together?
Thanks again.
Simon
The strains should still be normalized together.
Normalization reduces the between array variability. If the strains
are
normalized separately, the within strain variance will be falsely
deflated,
increasing the apparent significance of differential expression.
If the true distribution of probe expression differs dramatically
between
strains, then you are in real trouble. (I have seen this - not in
strains
but in treatments that directly affect the RNA.) I am not sure what
to do
- perhaps do quantile normalization using only the MM and control
probes,
and then normalizing the PM probes to the quantile distribution.
The microarray analysis community has not yet addressed the problem of
normalization when there is a lot of differential expression.
--Naomi
At 10:42 AM 4/15/2004, Simon Kidd wrote:
>At 8.50am -0400 15/4/04, James MacDonald wrote:
>>For your first question, it is almost always preferable to run rma
on
>>all samples in a set together rather than in two batches. For the
second
>>question, the bug only concerns justRMA and justGCRMA, so rma and
gcrma
>>are not affected.
>
>Thanks, however my collaborator thinks I did not phrase my question
very
>well and did not emphasis enough that the genetic background of the
>strains could be very different, so this is what she thinks the
question
>should be:
>
>
>>We have performed two experiments. One compares fly strain A with
>>mutation M to fly strain A with a duplication for the wild type gene
M.
>>The second experiment compares fly strain B with the same mutation M
with
>>a different duplication for the wild type gene M. Aside from the
>>particular mutation we are assaying the genetic background of the
strains
>>could be very different. If one normalizes arrays from different
>>tissues using dChip, it is suggested that the arrays get normalized
>>separately. Is this also true for RMA or should they be normalized
together?
>
>If the strains are very different should they still be normalised
together?
>
>
>Thanks again.
>
>Simon
>
>_______________________________________________
>Bioconductor mailing list
>Bioconductor@stat.math.ethz.ch
>https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor
Naomi S. Altman 814-865-3791 (voice)
Associate Professor
Bioinformatics Consulting Center
Dept. of Statistics 814-863-7114 (fax)
Penn State University 814-865-1348
(Statistics)
University Park, PA 16802-2111
Ah. In this case, since the two strains may be quite different, it may
be better to normalize separately.
Jim
James W. MacDonald
Affymetrix and cDNA Microarray Core
University of Michigan Cancer Center
1500 E. Medical Center Drive
7410 CCGC
Ann Arbor MI 48109
734-647-5623
>>> Simon Kidd <kidd@mail.rockefeller.edu> 04/15/04 10:42AM >>>
At 8.50am -0400 15/4/04, James MacDonald wrote:
>For your first question, it is almost always preferable to run rma on
>all samples in a set together rather than in two batches. For the
second
>question, the bug only concerns justRMA and justGCRMA, so rma and
gcrma
>are not affected.
>
Thanks, however my collaborator thinks I did not phrase my question
very well and did not emphasis enough that the genetic background of
the strains could be very different, so this is what she thinks the
question should be:
>We have performed two experiments. One compares fly strain A with
>mutation M to fly strain A with a duplication for the wild type gene
>M. The second experiment compares fly strain B with the same
>mutation M with a different duplication for the wild type gene M.
>Aside from the particular mutation we are assaying the genetic
>background of the strains could be very different. If one
>normalizes arrays from different tissues using dChip, it is
>suggested that the arrays get normalized separately. Is this also
>true for RMA or should they be normalized together?
If the strains are very different should they still be normalised
together?
Thanks again.
Simon
_______________________________________________
Bioconductor mailing list
Bioconductor@stat.math.ethz.ch
https://www.stat.math.ethz.ch/mailman/listinfo/bioconductor
