Transcript Microarray Gene Expression Data

MIAME and ArrayExpress –
a standard for microarray
gene expression data
and the public database at EBI
Susanna-Assunta Sansone
(Toxicogenomics project coordinator)
Microarray Informatics Team
EMBL- EBI (European Bioinformatics Institute)
Transcriptome Symposium, April 2002
CHU Pitié-Salpêtrière, Université Paris VI
Why have a public database?
 EMBL- EBI centre for research and services in
bioinformatics that makes and maintains public db:
•
EMBL Nucleotide Sequence, SWISS-PROT, Ensembl, MSD, etc.
 Practical reasons:
•
•
•
Easy data access
Resolves local storage issues
Common data exchange formats can be developed
 Scientific reasons:
•
•
•
•
Curation can be applied
Annotation can be controlled
Additional info can be stored that is missing in publications
Improve data comparison !
 Public standard can be applied
Talk structure
 MIAME standard
 MIAME annotation challenge:
• MGED BioMaterial Ontology
 Uses of MIAME concepts:
• ArrayExpress:
a public repository for gene expression data
• MIAMExpress
submission and annotation tool
Talk structure
 MIAME standard
Standard for microarray data Why?







Size of dataset
Different platforms - nylon, glass
Different technologies - oligos, spotted
References to external db not stable!
Array annotation
Sample annotation
Data sharing needs standardized way to
annotate and record the information!
Standard for microarray data MGED Group
 Microarray Gene Expression Data Group:
EBI + world’s largest microarray labs and companies
(Sanger, Stanford, TIGR, Universite D'Aix-Marseille II,
Affymetrics, Agilent, NCBI, DDBJ, etc.)
 MGED Group aims to
• Facilitate adoption of standards for:
– Experiment annotation
– Data representation
• Introduce standard for:
– Experimental controls
– Data normalization methods
General MIAME principles
 Minimum information about a microarray experiment
 NOT a formal specification BUT a set of guidelines
 Sufficient information must be recorded to:
• Correctly interpret and verify the results
• Replicate the experiments
 Structured information must be recorded to:
• Query and correctly retrieve the data
• Analyse the data
 MIAME- Brazma et al., Nature Genetics, 2001
• Sample source
• Sample treatments
• Extraction protocol
• Labeling protocol
Sample
MIAME
Hybridization protocol
Hybridisation
Array
• Array design information
• Location of each element
• Description of each element
• Image
• Scanning protocol
• Software specifications
• Quantification matrix
• Analysis protocol
• Software specifications
MIAME 6 parts of a microarray experiment
MIAME
Experiment
Sample
Hybridisation
Array
Sample
Hybridisation
Array
Sample
Hybridisation
Array
Sample
Hybridisation
Array
• Strategy
• Algorithm
• Control array elements
• 3 data processing levels
• Lack of gene expression
measurement units !
Normalisation
Final data
MIAME 6 parts of a microarray experiment
MIAME – Annotation challenge
 Annotation implementations are required !
• Avoid/reduce free text descriptions
• Use of controlled terms
• Definitions and sources for each term
• Remove of synonyms, or use of synonym
mappings
• Data curation at source (LIMS)
• Integration of controlled terms in query interfaces
 Facilitate data queries-analysis…….
A gene expression database from
the data analyst’s point of view
Genes
and
transcription units
Samples
Gene expression
matrix
Gene expression levels
A gene expression database from
the data analyst’s point of view
Genes
and
transcription units
Samples
• Array description:
- Gene annotations
• Sample annotations:
- Source
- Treatment
Gene expression
matrix
Gene expression levels
MIAME - Gene annotation
 Unambiguous identification
 Synonyms !
• Community approved names
• Alternative to gene names
 Usable external sources e.g.:
• EMBL-GenBank - sequence accession n.
• Jackson Lab - approved mouse gene names
• HUGO - approved human gene names
• GO categories - function, process, location
MIAME - Sample annotation
 Gene expression data only have a meaning in
the context of detailed sample descriptions !
 Usable external sources e.g.:
• NCBI Taxonomy - organisms
• Jackson Lab - mouse strains names
• Mouse Anatomical Dictionary – mouse anatomy
• ChemID – compounds
• ICD-9 – diseases classification
 More is needed…..
Annotation –
implementations required!
 Need an ontology to describe the sample:
• Defining controlled vocabularies and……
• ….Using existing external ontologies
 Integrate the ontology in LIMS and databases:
• Develop browser or interface for the ontology
• Develop internal editing tools for the ontology
 However some free text description is
unavoidable
Talk structure
 MIAME standard
 MIAME annotation challenge:
• MGED BioMaterial Ontology
What CV and ontology are?
 Controlled Vocabulary (CV):
• Set of restrictive terms used to describe
something, in the simplest case it could be a list
 Ontology is more then a CV:
• Describes the relationship between the terms in
a structured way, provides semantics and
constraints
• Capture knowledge and make it machine
processable
Sample annotation –
MGED BioMaterial Ontology
 Under construction by Chris Stoeckert (Univ. of
Penn.) and MGED members
 Use OILed (rdf, daml and html files available)
 Motivated by MIAME and guided by ‘case scenarios’
 Defines terms, provides constraints, develops CVs for
sample annotation
 Links also to external CVs and ontologies
 Will be extended to other part of a microarray
experiment that need to be described
Sample annotation –
MGED BioMaterial Ontology
an example
Sample source and treatment description,
and its correct annotation using
the MGED BioMaterial Ontology
classes and correspondent external references:
“Seven week old C57BL/6N mice
were treated with fenofibrate.
Liver was dissected out, RNA prepared………”
MGED BioMaterial Ontology
External References
Instances
©-BioMaterialDescription
©-Biosource Property
©-Organism
NCBI Taxonomy
©-Age
©-DevelopmentStage
Mouse Anatomical Dictionary
Mus musculus musculus id: 39442
7 weeks after birth
Stage 28
Female
©-Sex
©-StrainOrLine
International Committee on Standardized
Genetic Nomenclature for Mice
Charles River, Japan
©-BiosourceProvider
©-OrganismPart
C57BL/6
Mouse Anatomical Dictionary
Liver
©-BioMaterialManipulation
©-EnvironmentalHistory
©-CultureCondition
©-Temperature
22  2C
©-Humidity
55  5%
©-Light
12 hours light/dark cycle
©-PathogenTests
Specified pathogen free conditions
©-Water
ad libitum
©-Nutrients
MF, Oriental Yeast, Tokyo, Japan
©-Treatment
©-CompoundBasedTreatment
(Compound)
ChemIDplus
Fenofibrate, CAS 49562-28-9
(Treatment_application)
in vivo, oral gavage
(Measurement)
100mg/kg body weight
Talk structure
 MIAME standard
 Sample annotation:
• MGED BioMaterial Ontology
 Uses of MIAME concepts:
• ArrayExpress
a public repository for gene expression data
• MIAMEpress
submission and annotation tool
Uses of MIAME concepts
 Specifies the content of the information:
• Sufficient
• Structured
 Uses:
• Creation of MIAME-compliant LIMS or databases
e.g: ArrayExpress
• Development of submission/annotation tool for
generating MIAME-compliant information
e.g.: MIAMExpress
ArrayExpress – data flow
EBI
Web server
Users
Submission
Submission
LIMS
Browse-Query
MIAMExpress
ArrayExpress
MIAMExpress
Curation
database
Output
MAGE-ML
Update
Image
server
Central
database
Data
warehouse
ArrayExpress - details
 Implementation in ORACLE of the MAGE-OM model:
• Microarray gene expression - Object Model
• OMG approved standard (MGED and Rosetta, 2001)
• Model developed in UML
 Object model-based query mechanism:
• Automatic mapping to SQL
ArrayExpress
 Independent of:
• Experimental platform
• Image analysis method
• Normalization method
Central
database
Data
warehouse
 MAGE-ML data loader:
• Microarray gene expression - Mark-up Language generated from model
ArrayExpress – conceptual model
Experiment
Sample
Hybridisation
Array
Sample
Hybridisation
Array
Sample
Hybridisation
Array
Sample
Hybridisation
Array
Normalisation
Final data
MIAME 6 parts of a microarray experiment
ArrayExpress – simplified model
• Classes are represented by boxes
• Classes describe objects
• Related classes are grouped together in
packages
• MAGE-OM has 16 packages, ~ 150 tables
ArrayExpress data (via MAGE-ML)
Currently:
Near future:
• Human data - EMBL
•
(ironchip)
•
• Yeast data - EMBL
•
• S. pombe - Sanger Institute •
• Available as example
•
annotated and curated data
sets
•
Array descriptions - TIGR
Array description - Affymetrix
Mouse data - TIGR and HGMP
Anopheles data - EMBL
Direct pipeline - Sanger Institute
LIMS
Data - DESPRAD partners
• Toxicogenomics data- ILSI HESI
ArrayExpress – query interface
First release 12 Januray 2002
ArrayExpress –
link to Expression Profiler
External data, tools
pathways, function,
etc.
Expression
data
EP:PPI
Prot-Prot ia.
EP:GO
GeneOntology
EPCLUST
GENOMES
Expression data
URLMAP
provide links
sequence, function,
annotation
SEQLOGO
SPEXS
PATMATCH
visualise patterns
discover patterns
ArrayExpress – curation effort
 User support and help documentation:
• Ontologies and CV’s
• Minimize free text, removal of synonyms
• Help on MAGE-ML format and MAGE-OM
 MIAME compliance-check
 Curation at source (LIMS)
 To provide high-quality, well-annotated data
and allow automated data analysis
MIAMExpress - details
 Submission and annotation tool:
• Curators will monitor the submissions
 Based on MIAME concepts:
MIAMExpress
• Experiment, Array and Protocol submissions
• Generates MIAME-compliant information
 Uses MGED BioMaterial Ontology terms:
• Terms and required fields are explained
 Allows user driven ontology development:
• User can provide new terms and their sources
 Allows browsing:
• Array descriptions
• Protocols
MIAMExpress - details
 Version 1 launch in December 2002
 Expected users:
MIAMExpress
• Limited local bioinformatic support
• No LIMS on site
• Small scale users with custom made arrays
 Can be installed as local version:
• As a lab-book to annotate your experiment
• As part of a LIMS
 Interfaces:
• Version 1 is general
• Future versions, application specific interfaces
- Species specific
- Toxicogenomics specific (ILSI- HESI)
ArrayExpress - future
 Load public data into ArrayExpress:
• TIGR, EMBL, ILSI HESI, DESPRAD partners
 Improve query interfaces
 Launch MIAMExpress v.1 (Dec.2002)
 MIAMExpress v.2:
• Extended according to the user needs
• Integrated MGED ontology
• Increased usability, flexibility and scalability
 Develop curation tools
Acknowledgments
 Microarray Informatics Team at EBI (19 members):
•
•
•
•
•
•
Alvis Brazma (Team Leader and MGED President)
Helen Parkinson (Curation Coordinator)
Mohammad Shojatalab (MIAMExpress Database Programmer)
Ugis Sarkans (ArrayExpress Database development coordinator)
Jaak Vilo (Expression Profiler)
Curators and Programmers.
 MGED members and working groups:
• Alvis Brazma (MGED President, MIAME)
• Chris Stoeckert, U. Penn. (MGED Ontology Working Group)
Resources and ….messages
 Open sources resources:
•
•
•
•
ArrayExpress and MIAMExpress schema-access to code
MIAME document and glossary
MAGE-ML dtd and annotation examples
MGED Ontology and other resources………
www.mged.org / www.ebi.ac.uk/microarray
[email protected]
 Be aware of MIAME !
• Nature, Lancet and have already expressed their interest
• Founding agencies
 Join MGED meetings, tutorials and mailing lists:
• MGED-5 meeting in Japan (Sept. 2002)
• Ontology for BioSample description, EBI (Nov. 2002)