Organ Transplant Rejection on the Semantic Web

1
Organ Transplant Rejection on the Semantic Web

• Benjamin Good
• CIHR and MSFHR Strategic Training Program in
  Bioinformatics
• Department of Molecular Biology and Biochemistry
• Simon Fraser University
• Wilkinson Laboratory
• iCAPTURE Center,
• St. Pauls Hospital
• November 4, 2004

2
Research Questions

• Biological
• Why and how does the body sometimes reject
  transplanted organs? Specifically, what are the
  biomarkers of this process?
• Informatics
• Can knowledge that is geographically, socially,
  and conceptually distributed be represented such
  that connections and conflicts between facts
  from different domains can be identified
  automatically?
• Bioinformatics
• Can we computationally encode biological
  knowledge?
• Does this encoding help to answer biological
  questions?

3
Hypothesis

• By structuring biological information according
  to the principles of the Semantic Web, the
  scientific process can be made more efficient,
  more productive, and more enjoyable.
• The fundamental principles of the Semantic Web
• (as defined by me)
• Be (productively) lazy
• Take full advantage of what others have done.
• Share
• Make it possible for others to take full
  advantage of what you have done.

4
The World Wide Web
Web page
Hyperlink
Web page
Web page
Hyperlink
Hyperlink
Web page
Hyperlink

• One kind of relationship
• One kind of node
• Meant for human browsing

5
The Semantic Web
Has_location
Person
Has_name
Student
isa
Professor
isa
Has_advisor
6
Semantic Markup RDF
Animal
Mammal
Primate
Lemur
Human
Gorilla
7
Web Services

• Web Services are programs that can be executed by
  other programs connected via the internet.
• Services can produce, consume, and be described
  by semantically rich XML.
• Discovery
• Via service descriptions
• Registries now
• Service search engines next?
• Composition
• Complex services such as pipelines can be created
  by combining simple components.

8
(No Transcript)
9
Semantic Web Services
bioService
SequenceCompare
BlastP
BlastX
10
Semantic Markup RDF
Animal
Mammal
Primate
Lemur
Human
Gorilla
11
Approach to Allograft Rejection

• Informatics Build something new
• Use the technology of the Semantic Web to build
  an information resource centered around allograft
  rejection.
• Biology Use it to answer questions
• What.. use it to identify biomarkers (easy
  question?)
• Why use it to generate potential explanations
  (hard question!)
• Bioinformatics Show that building something new
  helped to answer harder questions faster.
• Analyze how the creation of the system influenced
  the process of the science.

12
Data Flow for Allograft Rejection

• Data from successful and failing transplant
  recipients
• Affymetrix gene chips
• iTRAQ quantitative proteomics
• Patient data

13
Why not just a database?
Gene Exp.1
Patients
iCAPTURE
Proteomics-2
Proteomics-1
QC
Patients
14
Open World Biology Needs Open World Informatics
iCAPTURE
15
The proliferation of data is Good! but only if
integration is taken seriously
16
Open World Informatics via Shared Ontologies
NCBI
Proteomics
NCI
Gene Exp
Patients
EBI
Gene Experiment Protein
GO
SNOMED
PathDB
17
More than a shared lexiconAdding Semantics

• OWL The web ontology language
• OWL is description logic that conveniently can be
  expressed with RDF-XML
• Description Logics are formal languages that can
  be used to make statements about concepts such
  as
• For all There exists
• DL can be coupled with inference engines to
  automatically
• Assign new instances to appropriate classes
• Check conceptual consistency
• Answer queries

18
An Example of Reasoning
Action
Tool
Add constraint
If transcription increases then so must
translation
ProtégéOWL
1)
RACER
Any Conceptual Disagreements?
No!
2)
Any Data/Instance Disagreements?
RACER
Yes!
3)
4)
Ack! Rule is not valid or there is an error in
the data, investigate both
19
Summary

• We plan to investigate the phenomenon of
  allograft rejection using techniques associated
  with the semantic web.
• In so doing we hope to test the hypothesis that
  conducting bioinformatics on the semantic web
  will benefit the practice of biology.
• Specifically we hope to
• Identify biomarkers associated with allograft
  rejection.
• Prove that the semantic web can be used to
  integrate biological data.
• Show that this integration effort benefits the
  study of allograft rejection and that the general
  principles of design used to achieve this benefit
  are generic enough to apply to many other
  biological domains.

20
Specific Tasks

• Learn more about allograft rejection and
  proteomics methods (iTRAQ).
• Define questions that the ontologies should be
  able to answer. (Collaborate!!)
• Identify existing ontologies that are pertinent
  to the questions.
• Develop new ontologies to fill in the gaps
  identified in 3. (Collaborate!!)
• Bind the data coming into iCapture to these
  ontologies (using web services)
• Apply a reasoning engine to check for consistency
  between the knowledge base and the data.

21
Thanks to the Wilkinson Lab
22
When I say collaborate with experts
iCAPTURE houses more than 200 researchers in
cardio-pulmonary disease and related fields.
Bruce McManus MD, PhD
23
Challenges

• Biological knowledge representation is hard.
• Having an accepted standard like OWL is good for
  integration, but
• Because of (1), other languages and other
  modeling approaches may be necessary. For
  example, probability may be necessary.
• Trust
• Providence
• The true benefits of the semantic web cant be
  seen until a critical mass of people decide to
  participate. Scary to have politics have so much
  influence over science.

24
Potential Value

• A global knowledge base
• Personal digital assistants with access to all of
  the worlds knowledge.
• A synthesis of data beyond the reach of the human
  mind.
• A synthetic long term and short term memory
  system for science.
• Emergence - from the multi-agent systems