Hi.
On Thu, Sep 18, 2008 at 5:10 AM, marco [EMAIL PROTECTED] wrote:
Dear All,
along a similar line I have the following question.
The standard for aroma.affymetrix (and other software) is to use as
reference the robust average across the sample.
I wonder if there is a rule of the thumb for how many arrays make a
good reference?
I'd say with 15-25 in-house arrays you do better than any number (I
tried 270) HapMap samples.
In the vignette 6 arrays are used. I guess this should be considered
enough?
Nope. The examples are using 6 arrays so they are relatively fast to
process. You should definitely aim for more if you want to use a
pooled reference.
Adding more arrays should in general improve the results?
Yes. Discussion the reference here, the noise level of the pooled
reference goes down. With twice as many arrays, the average/mean
array will have half the variance, that is, with four times the number
of arrays the standard deviation of the mean goes down to half.
The variance of the test arrays will eventually be much greater than
the variance of the average array. Then adding more arrays to the
reference will not make much differences.
In the specific case I have 6 arrays from stem cells to use for CN,
and I have also 20 arrays from another experiment (blod from healthy
people)
that I could add to get a more stable baseline. Would this be
reccomended?
I think so, especially if they were processed at the same microarray
facility and even more so if they were done close in time. Try with
and without and search for a deletion/amplification that you can spot
by eye. I would expect that you will see a much cleaner aberration
with the extra 20 arrays.
Cheers
Henrik
Cheers
Marco
On Aug 29, 8:38 pm, Henrik Bengtsson [EMAIL PROTECTED] wrote:
Hi,
On Fri, Aug 29, 2008 at 2:15 AM, mbaudis [EMAIL PROTECTED] wrote:
I can support Henrik's observation regarding the HapMap samples. In
the 500k setting, 25 HapMap produce more noise when used as reference
compared to our in-house data (comparable number; non-cancer samples).
When doing small pilot series of cancer samples in the 6.0 setting, we
needed a normal reference set. Again, HapMap did not lead to good
results (though here we still have to do in-house reference samples
for comparison).
thanks for sharing your observations. If more observe the same,
please report back so we can out rule a common use case.
Henrik: We all will appreciate a commented R batch, for a situation as
the one described above, showing your preferred way to do it ;-)
I put it on my t long todo list - I might have it ready by 2011 or so ;)
Ok, a more constructive comment is that anyone may contribute to the
documentation and provide their own vignettes. Please feel free to
write one up and we'll put it online. Then others will try the code,
find the bugs, ask questions, give feedback and it will iterate to
higher quality.
Practicalities: There was a time when a group members could edit any
page, but spambots signed up and messed it all up, so I had to remove
those edit privileges. Instead, one solution is to write up a Google
Document and publish it (make it readable to the public) and then we
can link to that page from the Google Group or alternatively
cut'n'paste the content. Also, there is a long time plan to migrate
to Google Sites and that allows us to embed external Google Docs as
is.
Cheers
Henrik
Michael Baudis.
--~--~-~--~~~---~--~~
When reporting problems on aroma.affymetrix, make sure 1) to run the latest
version of the package, 2) to report the output of sessionInfo() and
traceback(), and 3) to post a complete code example.
You received this message because you are subscribed to the Google Groups
aroma.affymetrix group.
To post to this group, send email to aroma-affymetrix@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/aroma-affymetrix?hl=en
-~--~~~~--~~--~--~---