Hi,
When comparing the results of RMA from the affy package and justPlier
from the plier package, I found that the summarized data are clustered
very differently. The cluster of the RMA results is very close to MAS5
values, showing 5 groups. However, PLIER results show a very different
and strange clustering pattern, almost like using single linkage. (See
attached figure)
I used Pearson correlation coefficient distance and average linkage
for
all clustering.
I remember attending a web talk about PLIER and the speaker mentioned
post-processing scaling in PLIER. Is that done in justPlier? Or is
this
an additional step that needs to follow justPlier? Or is it something
else?
Thank you for your help!
Here is my code:
> sessionInfo()
R version 2.4.0 (2006-10-03)
sparc-sun-solaris2.8
locale:
C
attached base packages:
[1] "tools" "methods" "stats" "graphics" "grDevices"
"utils"
[7] "datasets" "base"
other attached packages:
plier affy affyio Biobase
"1.4.0" "1.12.0" "1.2.0" "1.12.2"
> eset1 <- rma(myData)
> exprs(eset1) <- 2^exprs(eset1)
> eset2 <- justPlier(myData, normalize=T)
> exprs(eset2) <- 2^exprs(eset2)
Yiwen He
Generally plier shows good correspondence with RMA.
There are penalty parameters that tune the optimization of the
plier model. Some of them may make plier a bit more "RMA-like".
For instance, in my experiments, setting concpenalty=0.1 removes
a lot of the artifacts of different fold change between plier and RMA.
Still, the vast majority of the fold changes are similar -
running a density plot instead of a scatterplot shows it well.
-----Original Message-----
From: bioconductor-bounces@stat.math.ethz.ch
[mailto:bioconductor-bounces at stat.math.ethz.ch] On Behalf Of He,
Yiwen
(NIH/CIT) [C]
Sent: 09 November 2006 21:38
To: bioconductor at stat.math.ethz.ch
Cc: microarray
Subject: [BioC] results of justPlier vs. RMA
Hi,
When comparing the results of RMA from the affy package and justPlier
from the plier package, I found that the summarized data are clustered
very differently. The cluster of the RMA results is very close to MAS5
values, showing 5 groups. However, PLIER results show a very different
and strange clustering pattern, almost like using single linkage. (See
attached figure)
I used Pearson correlation coefficient distance and average linkage
for
all clustering.
I remember attending a web talk about PLIER and the speaker mentioned
post-processing scaling in PLIER. Is that done in justPlier? Or is
this
an additional step that needs to follow justPlier? Or is it something
else?
Thank you for your help!
Here is my code:
> sessionInfo()
R version 2.4.0 (2006-10-03)
sparc-sun-solaris2.8
locale:
C
attached base packages:
[1] "tools" "methods" "stats" "graphics" "grDevices"
"utils"
[7] "datasets" "base"
other attached packages:
plier affy affyio Biobase
"1.4.0" "1.12.0" "1.2.0" "1.12.2"
> eset1 <- rma(myData)
> exprs(eset1) <- 2^exprs(eset1)
> eset2 <- justPlier(myData, normalize=T)
> exprs(eset2) <- 2^exprs(eset2)
Yiwen He
--------------------------------------------------------
This email is confidential and intended solely for the use
o...{{dropped}}
Thanks!
Adding "concpenalty=0.1" does help when tested by hierarchical
clustering on arrays of the summarized data.
However ----
We are switching our machines from UNIX to Linux, and strangely, I
found
that the results from the two machines differ! The R and BioC versions
are all identical but results differ dramatically. In fact, when on
Linux, the results cluster very well even when concpenalty is set as
default.
Now, I also tested on PC and found that the results are identical on
PC
and Linux. Seems that there are some problems with the package run on
UNIX.
Has anyone had such a strange situation, with PLIER or any other BioC
packages?
Thank you very much!
Yiwen He
# UNIX:
> sessionInfo()
R version 2.4.0 (2006-10-03)
sparc-sun-solaris2.8
locale:
C
attached base packages:
[1] "tools" "methods" "stats" "graphics" "grDevices"
"utils"
[7] "datasets" "base"
other attached packages:
moe430acdf plier affy affyio Biobase
"1.14.0" "1.4.0" "1.12.0" "1.2.0" "1.12.2"
# Linux:
> sessionInfo()
R version 2.4.0 (2006-10-03)
x86_64-unknown-linux-gnu
locale:
LC_CTYPE=en_US.UTF-8;LC_NUMERIC=C;LC_TIME=en_US.UTF-8;LC_COLLATE=en_US
.U
TF-8;LC_MONETARY=en_US.UTF-8;LC_MESSAGES=en_US.UTF-8;LC_PAPER=en_US.UT
F-
8;LC_NAME=C;LC_ADDRESS=C;LC_TELEPHONE=C;LC_MEASUREMENT=en_US.UTF-8;LC_
ID
ENTIFICATION=C
attached base packages:
[1] "tools" "methods" "stats" "graphics" "grDevices"
"utils"
[7] "datasets" "base"
other attached packages:
moe430acdf plier affy affyio Biobase
"1.14.0" "1.4.0" "1.12.0" "1.2.0" "1.12.2"
# Windows:
> sessionInfo()
R version 2.4.0 (2006-10-03)
i386-pc-mingw32
locale:
LC_COLLATE=English_United States.1252;LC_CTYPE=English_United
States.1252;LC_MONETARY=English_United
States.1252;LC_NUMERIC=C;LC_TIME=English_United States.1252
attached base packages:
[1] "tools" "methods" "stats" "graphics" "grDevices"
"utils"
[7] "datasets" "base"
other attached packages:
moe430acdf plier affy affyio Biobase
"1.14.0" "1.4.0" "1.12.0" "1.2.0" "1.12.2"
-----Original Message-----
From: Michal Okoniewski [mailto:MOkoniewski@PICR.man.ac.uk]
Sent: Monday, November 13, 2006 11:25 AM
To: He, Yiwen (NIH/CIT) [C]; bioconductor at stat.math.ethz.ch
Subject: RE: [BioC] results of justPlier vs. RMA
Generally plier shows good correspondence with RMA.
There are penalty parameters that tune the optimization of the
plier model. Some of them may make plier a bit more "RMA-like".
For instance, in my experiments, setting concpenalty=0.1 removes
a lot of the artifacts of different fold change between plier and RMA.
Still, the vast majority of the fold changes are similar -
running a density plot instead of a scatterplot shows it well.
-----Original Message-----
From: bioconductor-bounces@stat.math.ethz.ch
[mailto:bioconductor-bounces at stat.math.ethz.ch] On Behalf Of He,
Yiwen
(NIH/CIT) [C]
Sent: 09 November 2006 21:38
To: bioconductor at stat.math.ethz.ch
Cc: microarray
Subject: [BioC] results of justPlier vs. RMA
Hi,
When comparing the results of RMA from the affy package and justPlier
from the plier package, I found that the summarized data are clustered
very differently. The cluster of the RMA results is very close to MAS5
values, showing 5 groups. However, PLIER results show a very different
and strange clustering pattern, almost like using single linkage. (See
attached figure)
I used Pearson correlation coefficient distance and average linkage
for
all clustering.
I remember attending a web talk about PLIER and the speaker mentioned
post-processing scaling in PLIER. Is that done in justPlier? Or is
this
an additional step that needs to follow justPlier? Or is it something
else?
Thank you for your help!
Here is my code:
> sessionInfo()
R version 2.4.0 (2006-10-03)
sparc-sun-solaris2.8
locale:
C
attached base packages:
[1] "tools" "methods" "stats" "graphics" "grDevices"
"utils"
[7] "datasets" "base"
other attached packages:
plier affy affyio Biobase
"1.4.0" "1.12.0" "1.2.0" "1.12.2"
> eset1 <- rma(myData)
> exprs(eset1) <- 2^exprs(eset1)
> eset2 <- justPlier(myData, normalize=T)
> exprs(eset2) <- 2^exprs(eset2)
Yiwen He
--------------------------------------------------------
This email is confidential and intended solely for the use
o...{{dropped}}
