MGED Ontology Workshop

1
MGED Ontology Workshop

• MGED7
• September 8-10, 2004
• Toronto, Canada

2
MAGE Workshop
May 24th- May 28th 2004 The Institute for
Genomic Research and University of Maryland,
Shady Grove Rockville, MD
3
MAGE Workshop Goals

• MAGE -gt MGED Ontology API
• Main goal for this meeting
• Build a mechanism for people to use the MO as
  part of MAGE
• Ontology Tools
• In order to use MO we need tools for manipulating
  and managing the ontology
• MAGE v2 Model
• Continue model discussions toward MAGE v2
• Document Model and changes as the model is
  developed
• Begin work on Mapping MAGE v1 to MAGE v2

4
MAGE Workshop Goals

• Standalone ADF Converter
• Continue work on simplified Array Design Format
  and a reader and writer for it
• Integrate this into MAGE v2
• Documentation
• MO policies and usage
• MOE class or methods
• MAGE v2
• Code, Code, Code

5
Ontology Tools

• daml file parsing scripts for extracting the
  classes, instances and properties from MO
• ANSI SQL scripts for creating MO in a relational
  database like MySQL or Sybase
• Script based methods for updating a datrabase
  implementation of the MO
• Perl and Java methods for searching the MO for
  classes, instances, and properties
• Others ?

6
OWL Reader

•   Adam Witney developed an reader based on the
  Redland RDF reader that will parse the MGED
  Ontology from .OWL as well as .DAML files. This
  reader became useful to the Ontology helper API.
  It is located in cvs lib/Perl/script/MGEDOntology
  _parser.pl

7
Ontology Helper API

• Eric Deutsch and Kjell Petersen developed
  MGEDOntologyHelper.pm which will create
  OntologyEntry objects based on leaf node data.
  Both applications follow MGEDOntology policies.
• MO Traversal application written by Kjell
  Petersen and Stathis Sideris.was extended to use
  leaf node data to instantiate an OE MAGE-ML
  object.
• Eric Deutsch ported Stathis code to a perl
  modules, which returns nested OE objects. The
  Perl helper is nearly complete. It works well in
  simple cases, but in nested cases, the final
  value doesnt get inserted.
• These prototype modules and applications are
  available in cvs
• Perl MAGE-Perl/MAG/Tools. Java
  MAGE-Java/MGEDOntologyEntry
• The code also is a working prototype.
• Example scripts are located in lib/Perl/MGEDOntol
  ogy
• The Jave Helper has been completed and is fully
  functional.

8

• Perl modules
• MGEDOntologyClassEntry.pm
• MGEDOntologyEntry.pm
• MGEDOntologyHelper.pm
• MGEDOntologyPropertyEntry.pm
• Java classes
• MGEDOntologyEntry.java
• MGEDOntologyClassEntry.java
• MGEDOntologyPropertyEntry.java
• OntologyHelper.java
• StringManipHelpers.java

9
Example Perl for MGEDOntologyHelper

• my ontologyEntry1 BioMAGEToolsMGEDOntolog
  yClassEntry-gtnew(
• parentObject gt qt, ref to a QT object
• className gt 'QuantitationType',
• association gt 'Scale',
• values gt
• Scale gt 'linear_scale',
• ,
• ontology gt ontology, ref to MO Helper obj
• )

10
Resulting OE

• ltOntologyEntry value"linear_scale"
• category"Scale"gt
• ltOntologyReference_assngt
• ltDatabaseEntry URI"http//mged.sour
  ceforge.net/ontologies/MGEDOntology.phplinear_sca
  le"
• accession"linear_scale"gt
• ltDatabase_assnrefgt
• ltDatabase_ref identifier"MO"/gt
• lt/Database_assnrefgt
• lt/DatabaseEntrygt
• lt/OntologyReference_assngt
• lt/OntologyEntrygt

11
Another example

• my ontologyEntry1 BioMAGEToolsMGEDOntolog
  yClassEntry-gtnew(
• parentObject gt BioSource, ref to parent
• className gt 'BioMaterial',
• association gt 'Characteristics',
• values gt
• OrganismPart gt 'lung',
• ,
• ontology gt ontology, ref to MO obj.
• )

12
Missing value in resulting OE

• ltOntologyEntry value"BioMaterialCharacter
  istics
  category"BioMaterialCharacteristics"gt
• ltOntologyReference_assngt
• ltDatabaseEntry URIhttp//mged.sourc
  eforge.net/ontologies/MGEDOntology.phpBioMaterial
  Characteristics accession"BioMaterialCharacteri
  stics"gt
• ltDatabase_assnrefgt
• ltDatabase_ref identifier"MO"/gt
• lt/Database_assnrefgt
• lt/DatabaseEntrygt
• lt/OntologyReference_assngt
• ltAssociations_assnlistgt
• ltOntologyEntry value"OrganismPart
  category"OrganismPart"gt
• ltOntologyReference_assngt
• ltDatabaseEntry
  URI"http//mged.sourceforge.net/ontologies/MGEDOn
  tology.phpOrganismPart"
• 
  accession"OrganismPart"gt
• ltDatabase_assnrefgt
• ltDatabase_ref
  identifier"MO"/gt
• lt/Database_assnrefgt
• lt/DatabaseEntrygt
• lt/OntologyReference_assngt
• lt/OntologyEntrygt

13
Autogenerated MO classes

• Scott Gustofsen has devised a method of
  generating Java classes from the MO. He calls
  this Java Ontology Bindings. The code isnt yet
  implemented, but will be in the near term. The
  classes would have to be regenerated with each
  release of MO. He has agreed to push it into
  cvs.

14
BioMoby Web service

• Tina Boussard and Derek Fowler began work to
  establish a BioMoby service to do the following
• 1) search MO for terms and definitions,
• 2) return instantiated classes objects,
• 3) provide command-line client register service
  for batch processes.
• The service would likely be hosted at CBIL
  (Upenn). A namespace has been defined datatype
  and output/input formats for searches still need
  to be defined. Currently, the service uses GO as
  the test case database. Derek Fowler has set up
  a BioMoby test service test_GetMoTerm.
• The Moby namespace was problematic because
  another group at Cornell has registered their
  namespace incorrectly. They've been asked to fix
  it.

15
MO Term Tracker/Validator

• Trish Whetzel implemented a term tracker for the
  RAD database.
• The use cases for the Tracker are
• 1) return new terms proposed by date and/or
  submitter,
• 2) return all terms for any MO class,
• 3) consistency check the MO.
• The Tracker would be hosted at CBIL. Helen
  Parkinson and Trish discussed methods and use
  cases for managing the MO in RAD.

16
ADF Converter

• Philippe Rocca-Serra continued work on the ADF
  format with Michael Miller and Pierre Bushel. In
  order to incorporate chip/cgh data he had to
  remove Reporter and CompositeSequence
  identifiers. The format now is in an Excel
  workbook with 3 worksheets headers, Reporters
  and CompositeSequences. The code will be placed
  in cvs when ready.

17
MAGEv2 Progress

• Ugis Sarkans, Paul Spellman, Michael Miller and
  Angel Pizarro continued re-model-ing discussions.
  The following changes are being considered
• 1. Use Channel to link FactorValues to particular
  BioMaterials
• 2. BioAssays are now generalized to (hopefully)
  include all types of experimental protocols. As
  such the BioMaterial Treatment object is now a
  type of BioAssay.
• 3. ArrayDesign and DesignElement will mostly be
  left as-is in the model and the reference
  implementation of the model will have native
  support for a simpler format, probably ADF. It is
  not yet clear whether the default serialization
  (XML schema) will have both formats.
• 4. Protocols have been changed to be multistep
  protocols, i.e. a set of ordered steps, and not
  just simple protocol descriptions. This has
  allowed the changes in BioAssay to take place.
• 5. A new abstract class, Referenceable was
  devised to separate the concept of internal MAGE
  references (Identifiable) to objects that exist
  in other resources (e.g. they have associations
  to DatabaseEntry and BibliographicReference)
• 6. OntologyEntry will be redefined to allow for
  representation of frame-based and descriptive
  logic ontologies (MO), as well as simpler
  node-based ontologies (GO)
• 7. HigherLevelAnalysis will be extended to
  represent other types of analytical results, The
  current cluster represnetation will remain the
  same, modulo some bug fixes.
• 8. The MAGE submitters notes were searched for
  best-practices issues that could be solved by the
  model. The conclusion was that most of the
  best-practice recommendations stem from semantic
  checks, not syntactic, so the model changes would
  not suffice.

18
MO Problems

• Kjell and Eric found a couple of inconsistencies
  in the Ontology which are being investigated -
  these have been posted into the MGED Ontology
  tracker.
• There was an issue about how we handle deprecated
  instances and classes this will also be
  investigated at the next VOW.

19
Ontology Helper API

• Eric Deutsch and Kjell Petersen developed an API
  which will create OntologyEntry objects based on
  leaf node data, i.e. the instance values for
  the OE. Both applications follow MGEDOntology
  policies.
• Kjell Petersen extended a Java application
  written together with Stathis Sideris after the
  EBI jamboree last December. The application
  traverses the MO to generate uninitalised data
  structures according to the policies, with the
  possible choices from MO available at each node.
  The extended code will now take a minimum set of
  'leaf node' data as input, validate them against
  MO and instantiate the full data structureof MAGE
  OntologyEntry. The code is a working prototype.
  The classes are in cvs MAGE-Java/MGEDOntologyEntr
  y. Kjell completed the Java Helper such that it
  produces complete and filled OE objects.
• Eric Deutsch ported Stathis code to a set of
  perl modules that are used by a further module,
  MGEDOntologyHelper.pm, which returns nested OE
  objects. The helper is nearly complete. It
  works well in simple cases, but in nested cases,
  the final value doesnt get inserted. These
  modules are available in cvs
• MAGE-Perl/MAGE/Tools. The code also is a
  working prototype. Example scripts that use it
  are located in lib/Perl/MGEDOntology as
  test1.pl , test2.pl, and test3.pl.).