Back to the Top
The following message was posted to: PharmPK
Dear All,
I have done a microarray study using six Affymetrix Rat genome 2.0 chips
(three for treated gp vs three for control gp).
I have the following questions and I`d appreciate if I can get the
input of
those who are experienced in microarray data analysis.
1-what are the criteria upon which genes can be selected?
-Fold change only regardless P-value (what is minimum acceptable fold
change?)
-P-value< 0.05 only regardless fold change
-Both, fold change and P-value (what is the cutoff value for fold
change
in this case?)
2-For genes to be filtered based on the calls (absent (A), present (P),
marginal (M)). How many present (P)/ marginal (M) calls per group are
enough to consider a gene present and suitable for subsequent
comparisons.
For example, if a study involved using 6 chips, three chips/group,
should a
gene have a presence call in
- all six chips composing both groups i.e P, P , P vs P, P , P
- only in three chips composing one group regardless of the
calls in the
other group i.e P, P, P vs A, M, A
or M, P, P vs A, M, A
- only in two out of three chips composing each group regardless
of the
calls in the other group i.e A, P, P vs A, M, A
3-what is the statistical test that should be used for comparing genes ?
Thanks in advance
Hazem E.Hassan
Ph.D. Candidate
Pharmacokinetics/Biopharmaceutics Laboratory
Department of Pharmaceutical Sciences
School of Pharmacy, UMB
20 Penn st, room 511
Baltimore,MD, 21201
410-706-7388
410-706-5017(fax)
hhass002.-at-.umaryland.edu
Back to the Top
The following message was posted to: PharmPK
Dear All,
I have done a microarray study using six Affymetrix Rat genome 2.0 chips
(three for treated gp vs three for control gp).
I have the following questions and I`d appreciate if I can get the
input of
those who are experienced in microarray data analysis.
1-what are the criteria upon which genes can be selected?
-Fold change only regardless P-value (what is minimum acceptable fold
change?)
-P-value< 0.05 only regardless fold change
-Both, fold change and P-value (what is the cutoff value for fold
change
in this case?)
2-For genes to be filtered based on the calls (absent (A), present (P),
marginal (M)). How many present (P)/ marginal (M) calls per group are
enough to consider a gene present and suitable for subsequent
comparisons.
For example, if a study involved using 6 chips, three chips/group,
should a
gene have a presence call in
- all six chips composing both groups i.e P, P , P vs P, P , P
- only in three chips composing one group regardless of the
calls in the
other group i.e P, P, P vs A, M, A
or M, P, P vs A, M, A
- only in two out of three chips composing each group regardless
of the
calls in the other group i.e A, P, P vs A, M, A
3-what is the statistical test that should be used for comparing genes ?
Thanks in advance
Hazem E.Hassan
Ph.D. Candidate
Pharmacokinetics/Biopharmaceutics Laboratory
Department of Pharmaceutical Sciences
School of Pharmacy, UMB
20 Penn st, room 511
Baltimore,MD, 21201
410-706-7388
410-706-5017(fax)
hhass002.aaa.umaryland.edu
Back to the Top
The following message was posted to: PharmPK
First:
#1 - adjusted p-value. Adjustment depends on nature of the data,
technical handling, and scientific issues. Fold change by itself is
completely useless, since variation can be condition-dependent.
#2 - Affy calls are better than random, but the usefulness seems to
apply randomly. You need to understand what exactly they mean by
"present" and "absent", the usual English definitions that those words
carry are only very crude approximations.
#3 - Augmented T-statistics (empirical bayes, SAM, or similar
augmentation).
You might consider reading through the vignettes on the BioConductor
WWW site for examples of processing and interpreting your data, along
with the code/instructions for doing it (http://www.bioconductor.org/,
and http://www.bioconductor.org/docs/vignettes.html ). However, it
could be that you don't have enough information (6 chips only?) get
anything other than random results. The BioC vignettes (PDF documents
combining data analysis and the code to generate the data analysis)
should step you through the best 20 or so approaches to processing the
affy chips. Not clear what the best are.
I know I'm sounding pessimmistic, there is a chance that you'll see
some thing useful and sensible, but in my experience, this sounds like
a seriously underpowered study, and if so, your results will be not
much better than rolling dice.
--
best,
-tony
blindglobe.-at-.gmail.com
Muttenz, Switzerland.
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)