1. Taverna and SoapLab
Elda Rossi – CINECA (Italy)

2. What is CINECA Cineca is a consortium of Italian Universities and CNR
Funded in 1969, now under the control of Research and University Ministry

3. Resources
The most important national infrastructure in Italy for the computational support to scientific research
Mission:
promoting the use of the most advanced computing systems to support public and private scientific and technological research

4. R & Bioconductor
Bioconductor is an open source and open development software project for the analysis and comprehension of genomic data.
It is based on R , a language and environment for statistical computing and graphics.

5. R & Bioconductor BioConductor is a collection of “packages”
Two main types:
provides basic infrastructure support.
Provides innovative methodology
We chose a function in the affy package (type 2. )

6. The affy package Package: affy
Description: The package contains functions for exploratory oligonucleotide array analysis. The dependance to tkWidgets only concerns few convenience functions. 'affy' is fully functional without it. Version: Author: Rafael A. Irizarry , Laurent Gautier , Benjamin Milo Bolstad , and Crispin Miller with contributions from …
Maintainer: Rafael A. Irizarry Dependencies: R (>= 1.9.0), Biobase (>= ), reposTools Suggests: tkWidgets (>= 1.2.2), affydata SystemRequirements: None License: LGPL version 2 or newer URL: None available
Function: Expresso . From raw probe intensities to expression values

7. The expresso function Expression measures
The most common operation is certainly to convert probe level data to expression values.
reading in probe level data
background correction 4 methods
Normalization methods
probe specific background correction, e.g. subtracting MM methods
summarizing the probe set values into one expression measure and, in some cases, a standard error for this summary 5 methods

8. How to run expresso Data.out Data.CEL Report script $ R
> library(affy)
> data<-ReadAffy()
> data.mas<-expresso(data,bgcorrect.method="mas",
pmcorrect.method="mas",
normalize.method="constant",
summary.method="medianpolish")
> write.exprs(data.mas,file=“Data.out")
Report
script
$ R CMD BATCH script
library(affy)
data<-ReadAffy()
data.mas<-expresso(data,bgcorrect.method="mas",
pmcorrect.method="mas",
normalize.method="constant",
summary.method="medianpolish")
write.exprs(data.mas,file=“Data.out")

9. The files OUT file CEL file [CEL] Version=3 [HEADER] Cols=126 Rows=126
TotalX=126
TotalY=126
Baseline=Not normalized
DatHeader=ctrl150:CLS=1167 …
[INTENSITY]
NumberCells=15876
CellHeader=X Y MEAN
Sample001.cel Sample002.cel Sample003.cel
100084_at
101482_at
31962_at
32466_at
35201_at
36189_at
36678_at
37001_at
37029_at
37046_at
37189_at
37719_at
37725_at
38437_at
38730_at
39425_at
40276_at
The CEL file describes the intensities determined for every feature on a chip, without providing information about which probes correspond to which probe sets (such information provided by the CDF or 1LQ file). This format is not a fixed standard, Affymetrix does not directly support file access and this format may change in the future. What we describe here is intended to explain CEL files in the context of this transcriptome data set only, not all of the statements here can be applied to CEL files for other Affymetrix chips. The file is a simple text file. There are a number of header and footer lines, the relevant data begins after the line containing the following text (line 24): CellHeader=X Y MEAN STDV NPIXELS and ends on the line containing the following text: [MASKS] The relevant lines are in tab-delimited format. There is a line for every probe on the chip, each line contains the following fields: field 1 = X, the x-coordinate of the probe. field 2 = Y, the y-coordinate of the probe. field 3 = MEAN, the 75th percentile of the pixel intensities in the stripped feature. field 4 = STDV, the standard deviation of the pixel intensities in the stripped feature. field 5 = NPIXELS, the number of pixels in the stripped feature. In the above, "stripped feature" means what remains after the outer layer of pixels has been stripped away.

10. Setting up SoapLab A linux based server was chosen
Tomcat was installed
Java was upgraded
Axis was installed
SoapLab was installed
Vega.cineca.it
Tomcat
Java 1.4
Axis 1.1
SoapLab precompiled for Suse Linux
Up to here: No Problems !!!

11. Defining the Application
Write the application wrapper
Write the ACD file for the application
Convert ACD to XML
Start up the SoapLab server
Deploy the new service

12. 1. Write the application wrapper
/biotools/services/affy-expresso.pl
# R code to run analysis
open(AFFY,">$datadir/affy");
print AFFY <<EOF ;
library(affy)
data<-ReadAffy()
data.mas<-expresso(data,
bgcorrect.method="$bgcorrect",
pmcorrect.method="mas",
normalize.method="$normalize",
summary.method="medianpolish")
write.exprs(data.mas,file="data.txt")
EOF
close(AFFY);
# now run program
system "cd $datadir; $rexe CMD BATCH affy";
# print output
open(OUT,"$datadir/data.txt");
while (<OUT>) {print $_;}
close(OUT);
#!/usr/bin/perl
use Getopt::Long;
# command arguments (with default)
GetOptions("bgcorrect=s"=>\$bgcorrect,
"normalize=s"=>\$normalize);
$bgcorrect="mas" if $bgcorrect eq "";
$normalize="constant" if $normalize eq "";
# location of R executable
$rexe="/biotools/R/R-2.1.0/bin/R";
# data directory
$datadir=“/biotools/services/data";

13. 2. Write the ACD file
/biotools/soapbin/analysis-interfaces/metadata/affy.acd
appl: bioconductor [
documentation: "affy/expresso function of BioConductor"
version: "1.0"
groups: "Microarrays"
nonemboss: "Y"
executable: affy-expresso.pl ]
string: bgcorrect [
additional: "Y"
parameter: "Y"
default: "mas"
string: normalize [
additional:"Y"
default: "constant"
outfile: output [
default:“stdout"
The path is defined in the shell
Input1: Background correction
Input1: Normalization method
Output: standard output

14. 3, 4, 5: Final steps Convert ACD to XML Start up the SoapLab server
Deploy the new service
/biotools/soapbin/analysis-interfaces/generator/acd2xml
From: ../metadata/affy.acd
To: ../metadata/microarrays/affy-al.xml
/biotools/soapbin/analysis-interfaces/run-AppLab-server
How to shut down the server?
/biotools/soapbin/analysis-interfaces/ws/deploy-web-services

15. Using the service from Taverna
From the Available service window select Add new SoapLab scavenger and enter our server address

16. Using the service … (2) The new processor appears
in the microarrays folder you can find the affy service
After connecting input & output ports, the service can be launched

17. Problems encountered Documentation is not so clear and complete
How can we transfer (large) files from the personal WS to the server machine
We need a permanent and private data area for storing data
We would like to monitor the service while it is running, batch support (asynchronous services?)
How can we return data in addition to stdOut and stdErr
…..

18. A possible (future) workflow
Upload one or more CEL files on the server
WS-upload
Analyse the data and get expression levels
WS-expresso
WS-plot
Verify the output data
YES NO
OK ?
download the output data and clear the personal space