An Ontology for ProteinProtein Interaction Data

1
An Ontology for Protein-Protein Interaction Data

• Karen Jantz
• CIS Honors Project
• December 7, 2006

2
Overview

• Problem Statement
• Objectives
• Approach
• Background
• Methodology
• Evaluation
• Demonstration
• Conclusion

3
Problem Statement

• Several sources for protein-protein interaction
  data
• Different schemata
• Different purposes
• Different strengths/weaknesses

4
Objectives

• Unify the data
• Enable data mining
• Evaluate reliability of data across data sources
• Gain new information about the entire data set
• Enable others to easily add other data sources to
  the set

5
Approach ontology

• ontology n.
• that which exists (philosophy)
• that which is represented (artificial
  intelligence)
• A descriptive data model
• Defines the entities and relationships within a
  domain
• Based upon data
• Human-readable

6
Approach ontology

• Data integration
• Enables simultaneous querying across multiple
  databases
• Data transformation
• Enables interchange between database formats
• Data mining
• Enables reasoning and learning over the entire
  data set

7
Background Data Sources

• DIP (Jing Xia)
• Database of Interacting Proteins
• Most reliable data set
• Jing Xia
• BIND (Abhijit Erande, Aaron Schoenhofer)
• Biomolecular Interactions Network Databank
• Very large data set
• Contains interactions, molecular complexes, and
  pathways

8
Background Data Sources

• MINT
• Molecular INTeractions database
• experimentally verified protein interactions
• Evaluates confidence level
• IntAct
• Not limited to binary interactions
• Allows user submissions
• mips CYGD
• Munich Information Center for Protein Sequences
  Comprehensive Yeast Genome Database
• Limited to yeast
• Focuses on sequencing

9
Background Tools

• Protégé
• Open-Source Project
• Graphical ontology editor
• Interacts with OWL Reasoner
• Detailed API for modifying ontologies
  programmatically

10
Background Tools

• Prompt
• A Protégé Plugin
• Enables ontology mapping
• Enables ontology comparison

11
Background Related Work

• PSI-MI
• Controlled vocabulary for PPI data
• Not a proposed database structure
• Decreases the strength of information
• Helpful in defining relationships and keys

12
Methodology Overview
Web Interface
Q What interactions have been observed between
with protein A?
Q What experiments give evidence for a given
interaction?
Unified Ontology
Unified Data Set
transformation
DIP
BIND
MIPS
MINT
IntAct
13
Methodology Design

• Review the singular database schemata and
  determine strengths/weaknesses
• View data files
• Native formats
• PSI-MI formats
• Create a unified schema of the data sources
• Create the unified ontology in Protégé
• Create each singular database as a subset of the
  unified ontology

14
Protégé Screenshot
15
Methodology Data Import

• DOMParser
• Load data from XML
• Protégé-OWL API
• Insert entities into singular databases

16
Methodology Transformation

• Use Prompt to create a mapping for each specific
  data source to the unified ontology
• Use Prompt mappings to insert individuals from
  each singular ontology into the unified model

17
Methodology Transformation

• Duplicate Data
• Need to fill in attributes on existing records
• Write Algorithm Plugin for Prompt to determine
  when individuals are the same

18
Prompt Screenshot - Mapping
19
Methodology Query Interface

• Export Protégé data into MySQL
• Web interface for collecting data
• Working with domain experts to determine useful
  views, queries

20
Evaluation

• Performance
• Transformation Time in Protégé
• Query Time for Web Interface
• Size
• Minimize redundancy in data model
• Minimize duplicate data

21
Evaluation

• Correctness
• Domain Experts
• Dr. Brown, Dr. Wang
• Maintain proper data relationships
• Utility
• Enrich data

22
Evaluation
23
Demonstration
24
Future Work

• Complete transformations
• Import data
• Evaluate ontology
• Add other databases to model

25
Conclusions

• Adequate start
• Needs improvement, evolution, more data sources
• As the project matures, the ontology will be
  ready for use in the biological domain
• Will be able to more easily gain information
  about protein-protein interactions

26
References

• AAAI.org - AITopics Ontology
• http//www.aaai.org/AITopics/html/ontol.html
• Protégé
• http//protege.stanford.edu/overview/protege-owl.h
  tml
• Prompt
• http//protege.cim3.net/cgi-bin/wiki.pl?Prompt
• PSI-MI
• http//psidev.sourceforge.net/mi/xml/doc/user

27
References

• BIND
• http//www.bind.ca
• DIP
• http//www.dip.doe-mbi.ucla.edu
• IntAct
• http//www.ebi.ac.uk/intact/site/
• MINT
• http//mint.bio.uniroma2.it/mint/Welcome.do
• MIPS
• http//mips.gsf.de/genre/proj/yeast

28
Q A