EMERALD Ontology Efforts

1
EMERALD Ontology Efforts

• James Malone, Helen Parkinson

2
Overview

• Background
• Ontologies the what and why?
• Building the best ontology we can Global
  collaboration
• Progress to date
• Future Steps

3
Background The Trouble with Microarray data

• Successful analysis and reproducibility of
  microarray experiments is dependant upon quality
  documentation and descriptions that are used to
  report microarray experiments.
• MIAME (Minimum Information About a Microarray
  Experiment)
• MAGE-TAB (MicroArray Gene Expression Tabular)
• Reproducing experiments requires reporting of
  transformations on data that are
• Unambiguous
• Consistent
• Understandable (in use of language and context)

!
4
A Common Question What on Earth are ontologies?

• Originally a philosophical invention from ancient
  Greece
• Used to describe the entities of existence and
  their relationship within this framework
• For our purposes we will use the widely cited
  definition of Tom Gruber An ontology is a
  specification of a conceptualization.
• In other words, they describe explicitly the
  concepts, the objects and relationships that hold
  among them that exist in some given domain.
• This last point is important because we are
  defining the world within some specific scope,
  e.g. gene expression or genetics or biology or
  science generally.
• Clearly the larger the scope or the community,
  the more complex the domain to model (and the
  more people have to agree on our model
  definitions!) this can make ontological
  modelling hard

challenging
5
Why do we need an ontology?

• Consider some of the reasons as to why anyone in
  bioinformatics uses ontologies
• Semantics The meaning of meaning?
• Ontologies define the syntax and semantics of
  concepts and relationships that hold between
  these concepts for a given domain richer
  representation of data
• Information sharing shared understanding
• Explicitness helps to remove ambiguity and helps
  other understand what it is we mean
• Machine readable (To Computer Scientists the most
  interesting part ?)
• Using languages such as Web Ontology Language
  (OWL), ontologies can be interpreted by software
  programmes

6
Why do we need an ontology in EMERALD?

• Diversity in microarray experiment designs and
  applications requires that a large number of
  pre-processing approaches are available
• Our previous check list
• reporting of transformations on data that are
• Unambiguous
• Consistent
• Understandable (in use of language and context)
• Powerful querying of biological models

7
An Ontology Example Querying and Browsing
The Arabidopsis Information Resource
8
An Ontology Example Visualisation and
MappingE.g. Edinburgh ATLAS
9
How Are We Building the Normalisation and
Transformation Ontology (NTO)?

• NTO is a coordinated action involving members of
  an assembled working group from around the globe
• Including biologists, biochemists, ontologists,
  computer scientists, statisticians, phillosophers
  and MDs.
• Collaboration with OBI project (more in a minute)
• Weekly teleconference calls
• Use of SVN for version controlling of ontology
• Face to face workshop meeting with working group
• Encouraging submission from potential users
• Dissemination

10
Ontology for Biomedical Investigations
(OBI)http//obi.sourceforge.net/

• OBI has the grand scope of enabling the modelling
  of any biomedical investigation, regardless of
  domain
• Orthogonal coverage, reuse of existing resources
  and shared frameworks

Cell Type Ontology
Chemical Entities of Biological Interest (ChEBI)
OBI
11
NTO Progress to Date

• Use case collection (recently started to place
  some online https//wiki.cbil.upenn.edu/obiwiki/in
  dex.php/EvaluationPhase1Submissions )
• Still welcoming submissions to (malone_at_ebi.ac.uk
  or Obi-datatrfm-branch_at_lists.sourceforge.net)

Feedback and iterate
Building
Identify Scope
Capture
Coding
Integrating
Evaluation
Version 1.0
Version 1.1
Document
12
Use Cases and Competency Questions

• Competency questions
• Which genes have a 2 fold change in expression
  where MAS5 has been applied as a data
  transformation methodology?
• Which pre-processed microarray data expresses
  values as log ratios (of two conditions) for a
  specified logarithmic base?
• Use Case
• An experimenter has conducted an expression
  microarray experiment involving two conditions
  with replicate assays per condition, where they
  have both biological and technical replicates.
  They are running two kinds of differential
  expression analyses (a) one at the gene level
  and (b) one at the gene set level. In (a) the aim
  is to identify differential expressed genes (e.g.
  via algorithms like PaGE and SAM). In (b) the aim
  is to identify, from an a priori given collection
  of gene sets (e.g. user provided, or based upon
  GO annotation), which of these sets are
  differentially expressed as a whole (e.g. via
  algorithms like GSEA or SAM-GSA). Before running
  the analyses the data is preprocessed with the
  following data transformation series (i) filter
  out flagged reporters, (ii) normalize the
  individual assays, (iii) average across technical
  replicates (but not across biological
  replicates). The above steps all requires
  annotation using the ontology.

13
NTO Progress to Date
14
NTO Progress to Date
15
Conclusion

• An NTO would give us
• Consistency in usage of terms through explicit
  definitions
• Widen reproducability of microarray experiments
• Richer representations, again definitions, but
  also axioms, relationships, properties to
  describe the data
• Reduction of disparate efforts
• (potentially) mappings to external resources
• Most importantly the biology and the data are
  given more relevance and increased utility
• But relies on collaborative efforts and
  consensus of opinion across domain (not always
  easy!)
• Annotating or modelling data with the ontology,
  i.e. actual use

16
Future Steps

• Meeting with OBI consortium and NTO working group
  member in January in Vancouver, Canada
• Publishing of Alpha version of NTO integrated
  within OBI early 2008.
• Evaluation of Alpha version with competency
  questions and use case
• Run annotation using ArrayExpress as assessment
• Iterate!!

17
Thanks

• EMERALD Consortium (www.microarray-quality.org)
• OBI Consortium (http//obi.sourceforge.net/)
• Especially Tina Boussard, Ryan Brinkman, Melanie
  Courtot, Elisabetta Manduchi, Monnie McGee, Helen
  Parkinson, Philippe Rocca-Serra, Richard
  Scheuermann
• ArrayExpress team (www.ebi.ac.uk/arrayexpress/)
• All contributors and submitters to ontology