1. Developing a protein-interactions ontology
Esther Ratsch
European Media Laboratory

2. PIOG Protein Interactions Ontology Group Computer scientists:
Philipp Cimiano Lavin (IMS Stuttgart, EML Heidelberg)
Isabel Rojas (EML Heidelberg)
Computational linguists:
Uwe Reyle (IMS Stuttgart)
Jasmin Saric (EML Heidelberg)
Biologists:
Esther Ratsch (EML Heidelberg)
Jörg Schultz (MPI for Molecular Genetics Berlin)
Ulrike Wittig (EML Heidelberg)

3. Motivation Why protein interactions? Why an ontology?
protein function analysis
larger datasets
Why an ontology?
clear domain model
storage and understanding of data
information retrieval from text
retrieve hidden information, inferencing

4. What is a signal transduction pathway?
signal from outside is transduced to the nucleus
often phosphorylation cascade
change
signal
transcription

5. Why are they important? control of cellular processes
communication between cells
response to environmental changes
regulatory network
stable system, single mutations may be overriden by other pathway
complex network enables complex behaviour

6. Jak-Stat pathway ligand cytokine receptors JAKs nucleus
STAT monomers P P
nucleus P P
tyrosine residues
target genes

7. Scope Concepts Representation Rules/Constraints Formalisation
General approach
Identify scope of the ontology
Identify concepts involved and their properties
How to represent them?
Define rules and constraints
Formalisation
Scope Concepts Representation Rules/Constraints Formalisation

8. Scope Concepts Representation Rules/Constraints Formalisation
The scope
Ontology that represents interactions between proteins and other cellular compounds
Restriction on molecular detail: amino acids
Concentration on signal transduction pathways in initial phase
no quantitative properties are modeled
Scope Concepts Representation Rules/Constraints Formalisation

9. Identify concepts: Interacting compounds
Different kinds of compounds: proteins, genes/DNA, ions, ...
Composition of compounds, e.g. amino acids, domains
DNA region
Jak
Stat
TAD LZ
DBD
SH3
SH2 Y
Domain organisation
of Stat proteins
Scope Concepts Representation Rules/Constraints Formalisation

10. Properties of compounds
Characteristics: molecular weight, sequence, isoelectric point...
Interaction potential: modifications, location, binding partners
nucleus P X
Scope Concepts Representation Rules/Constraints Formalisation

11. Identify concepts: Interactions I
Different types of interactions: phosphorylation, binding, translocation ...
Other classification: grouping of > 100 verbs (Swissprot)  11 not disjoint classes
Control/Regulation
Biochemical Interactions
Logical Interactions
Bind/Dissociate
Formation
Integrity
Availability
Change of Location
Modification of Structure
Special Processes/ Reactions
Order
Scope Concepts Representation Rules/Constraints Formalisation

12. Representation of proteins
General characteristics: sequence, molecular weight, ...
Protein state:
location
list of modifications
list of binding partners
StatState3(cytoplasm, phosphorylatedAtResidue701, Stat) P
JakState1(cytoplasm, none, cytokine-receptor)
Scope Concepts Representation Rules/Constraints Formalisation

13. Representation of interactions
Event with pre- and postconditions
not p
event e p t
phosphorylation P
not phosphorylated
phosphorylated
Scope Concepts Representation Rules/Constraints Formalisation

14. Scope Concepts Representation Rules/Constraints Formalisation
Rules and constraints
Simple hierarchies: nucleolus inside nucleus, Stat1 is a Stat is a protein
Rules for the definition of interactions
Consistency checking
Knowledge retrieval
Scope Concepts Representation Rules/Constraints Formalisation

15. Rules and constraints: example
„Protein A is phosphorylated by B at position X.“
A and B are located in the same compartment
A was not modified at X before
A is phosphorylated at X afterwards
B is a protein kinase, which is a protein
dependent on X, B is either a S/T-kinase or a Y-kinase
Scope Concepts Representation Rules/Constraints Formalisation

16. Formalisation: phosphorylation
Phosphorylation of a protein by a kinase at a distinct residue
S/T-kinase phosphorylation
Scope Concepts Representation Rules/Constraints Formalisation

17. Challenges met Multidisciplinarity of the group Domain
Different vocabularies  clear expression, fewer ambiguities
Different goals, different needs  not restricted to one goal
Different experiences  mutual benefit
Domain

18. Complexity of the domain
Granularity of information
detail of compound part
protein: Stat
domain: SH2-domain
amino acid: tyrosine701
detail of protein identity
protein family: Jak, Stat
protein type: Jak2, Stat5
organism specific protein: Jak2_human, Jak2_rat

19. Complexity of the domain II
description detail:
not known: no data available
doesn‘t have: no binding partners
don‘t care: not important for a certain interaction

20. What comes next? Go on with development of ontology
Projects using the ontology:
integration in larger ontology on metabolic pathways
application to TIGERSearch (see poster)

21. Acknowledgements Protein Interactions Ontology Group
Computer scientists:
Philipp Cimiano Lavin (IMS Stuttgart, EML Heidelberg)
Isabel Rojas (EML Heidelberg)
Computational linguists:
Uwe Reyle (IMS Stuttgart)
Jasmin Saric (EML Heidelberg)
Biologists:
Esther Ratsch (EML Heidelberg)
Jörg Schultz (MPI for Molecular Genetics Berlin)
Ulrike Wittig (EML Heidelberg)