MAGE and ArrayExpress

1
MAGE and ArrayExpress

• Ugis Sarkans
• EBI

2
ArrayExpress Infrastructure
Local MIAMExpress Installations
EBI
Submissions
www
MIAMExpress (MySQL)
Array Manufacturers
MAGE-ML
Queries
Datapipelines
ArrayExpress (Oracle Tomcat)
www
MAGE-ML
LIMS
Data analysis
www
Data Analysis software
Expression Profiler
MAGE-ML import, export
Microarray software
External Bioinformatics databases
Other Microarray databases
3
ArrayExpress architecture
Web page template
Web page template
MAGE-ML (DTD)
Tomcat
Velocity template engine
MAGE-ML (doc)
MAGE-OM
MAGE-ML (doc)
MAGE-ML (doc)
Java servlets
MAGE loader
MAGE validator
MAGE unloader
Castor
object/ relational mapping
ArrayExpress (Oracle)
error.log
4
AE schema

• Generated from (modified) MAGE-OM
• Classes ? tables
• Attributes ? fields
• 1..1, 1..n associations ? foreign keys
• m..n associations ? link tables
• Superclasses with incoming associations ? special
  tables supporting object-relational layer
  (Castor)

5
AE schema

• Some say that this cannot work, but
• AE must be able to import any valid MAGE-ML and
  not lose information
• good for dumping in MAGE-ML
• good for navigating through data in terms of
  object model
• if some queries dont work well, add something to
  the schema
• Experiment-Biomaterial, Experiment-Protocol links
• so far works for 80Gb of data

6
ArrayExpress purpose

• main objective - help in finding and initial
  exploration of data download for detailed
  analysis
• data repository (now) data warehouse (in
  development)

7
Queries - logical structure
Organisation - name
Person - last name
Array Design - accession - name
Array
Experiment - accession
Hybridisation
Species
Sample
ExperimentType
Protocol Type
ExperimentDesign
ExperimentalFactor
Protocol - accession
8
ArrayExpress other technical details

• Data matrices - stored in NetCDF format
• binary format for efficient storage of
  multidimensional array
• Arrays - stored as ADF spreadsheets (in addition
  to normal MAGE structures)

9
Data representation
in MAGE/ArrayExpress
in typical data analysis programs
10
ArrayExpress data export
BioAssayData2
11
Data export form
12
Array representation - ADF format
13
More complicated queries

• retrieve all experiments that use S.cerevisiae,
  in which any of the genes in my favorite pathway
  were down-regulated
• retrieve all experiments performed on mice, in
  any of the three tissues a, b, or c, where the
  fold change for gene x is greater than 2.0.

14
ArrayExpress DataWarehouse - very simplified view
Design element/ gene
Sample
BioAssay (hyb)
Array design
Experiment
15
AE DW - some more details
gene name
Properties
Properties
Properties
bioassay type
Properties
Properties
Properties
16
Next Jaboree

• EBI, Hinxton (close to Cambridge, UK)
• December 1 to 5/6, 2003
• Main objective ensure that MAGE-ML that comes
  from various sources is comparable
• Registration fee (for cost recovery)
• Participation mostly by invitation

17
Acknowledgements

• Gonzalo Garcia Lara - web interface
• Ahmet Oezcimen - DBA
• Anjan Sharma - curation tool
• Niran Abeygunawardena - webmaster
• Curation team
• Helen Parkinson, Philippe Rocca-Serra, Ele
  Holloway, Susanna Sansone, Gaurab Mukherjee
• Expression Profiler team
• Jaak Vilo, Misha Kapushesky, Patrick Kemmeren
• Alvis Brazma