Transcript Pelig slides

Integrating bio-ontologies with a
workflow/Petri Net model to qualitatively
represent and simulate biological systems
Mor Peleg, Irene Gbashvili, and
Russ Altman
Stanford University
Components of a biological model
Sequence components
Alleles, mutations
DB entries
Gene products
Cellular location
Molecular function
Proteolysis
Transport
Gene regulation
Biological process, clinical phenotype
Goals
• Piece together biological data
• Develop a qualitative model at first
–Data is noisy and incomplete
• Create a quantitative model eventually
• Store knowledge to allow
–systematic evaluation by scientists
–input for computer algorithms
Desired properties of a biological
processes model
• Represent 3 aspects of a biological system
– Molecular structures, functional roles, processes dynamics
•
•
•
•
•
•
Include a bio-medical ontology (concept model)
Display information graphically
Support hierarchical decomposition (complexity)
Provide formal semantics to verify correctness
Simulate system dynamics
Answer biological queries (reasoning)
– Proteins with same substrates, scoped to cellular location
– Alleles with roles in dysfunctional processes & disorders
Do other models posses the
desired properties?
Model
graf
nesting
static
function
dynamic
bio info
verify
Simulati
on tools
Computatio
nal model
GO
+
+
-
TAMBIS
+
+
DL
EcoCyc
+
+
+
+
frames
+
+
+
frames
Rzhetsky
+
PIF/PSL
+
I
KIF
BPML
+
C
XML
Workflow
+
+
Statecharts
OMT/UML
+
+
+
+
+
+
OPM
+
+
+
+
Petri Net
+
+
our model
+
+
C= components, I = integrated
+
+
I
+
+
+
Petri Nets
+
statechart
C
+/-
statechart
I
+/-
Semiformal
+
+
Petri Nets
+
+
Petri Nets
C
I
+
+
I
+
System Architecture
Biological Process Model
Workflow Model
Biological
data
Dynamic
data
Functional
data
Process Model
Organizational Model
OPM
Biomedical Ontology
Structural Data
TAMBIS
UMLS
Framework
developed in
Protégé-2000
Extensions
Petri Nets
Mapping business workflow to
biological systems
Business Workflow model
Biological Process Model
Process model
Process model
Organizational model
Structural model
Organizational Unit
(Faculty)
Biomolecular complex
(Replication complex)
member
member
Human
Role
(Dean)
Biopolymer
(Helicase)
(mapped
to TAMBIS)
Role
(DNA unwinding)
Systems modeled
• Malaria
Peleg et al., Bioinformatics
18:825-837, 2002
• Translation
Peleg et al., submitted to
P IEEE
Protein translation
aa1 aa2
aa3
aa7
aa4
aa5
E
G
aa6
P
A
U
Process Model: translation
elongation
E
tRNA0
P
tRNA1
A
tRNA0
tRNA1
tRNA2
tRNA1
tRNA2
tRNA1
Low level
Process
High level
Process
Check
point
Participant
tRNA2
process flow
substrate
product
affect
participation
inhibition
Other extensions
• Alleles and mutations
• Nucleic acid 2° and 3 ° structure
tRNA mutations affect translation
Frame-shifting
Misreading
Halting
aa1 aa2
aa3
aa7
aa4
aa5
E
G
aa9
aa6
P
A
U
Participant-Role Diagrams
Participants Relations
Individual
molecule
Complex
Complex-subunit
Collection-participant
Molecule-domain
Collection
Roles
Functional
role
Disease
role
<role>
specialization
role
Queries
Mapping to Petri Nets
van der Aalst (1998). The Journal of Circuits, Systems and Computers 8, 21-66
tRNA0 in E
site
P
1`a
Transient
binding to A
1`a
tRNA0 in E
P, A occupied
1`b
tRNA1 in P
E A occupied
1`a
1`b
tRNA0
exits
1`b
tRNA1 in P
A occupied
1`a
P
P
tRNA2 in A
E, P occupied
1`c
Binding to
A-site
1`c
tRNA2 in A
A occupied
1`c
P -> E
A -> P
1`b
1`c
1`b
Val_tRNA
Leu_tRNA
Phe_tRNA
1`c
1`b
tRNA1 in E
P occupied
Free tRNA
tRNA1 in P
site
1`b
tRNA2 in
1`c
Ternary
P
tRNA2 in P
E occupied
1`c
[(c<>Terminator_tRNA) and
(c<>Lys_Causing_Halting)]
1`b
Ready to
bind
1`c
Simulating abnormal reading
tRNA1 in P
site
tRNA0 in E
site
a
[c1]
Reading
b
[c2]
Misreading
tRNA2 in
Ternary
c
[c4]
[c3]
Frame shifting
Normal
current aa
Halting
Mutated
current aa
tRNA0 in E
P, A occupied
tRNA1 in P
E A occupied
tRNA2 in A
E, P occupied
[c2] = [(c = Misreading_tRNA)]
We also have places for nucleotides of current codons that feed in to the reading transitions
[c2] = [(c = Misreading_tRNA) and (x= C) and (y = C) and (z = C)]
Usefulness of Petri Nets
• Representing states explicitly
• Verifying dynamic properties (Woflan)
–liveness, boundedness
• Simulating dynamic behavior (Design/CPN)
• Reasoning on dynamics
–When inhibiting an activity, will we still reach a
certain state?
–Do competing models have different dynamics?
» Models of translation have different dynamics
Conclusion
• Our work integrates and extends three
unrelated knowledge models, enabling:
–representation of 3 aspects of biological systems
–reasoning on relationships among processes,
participants, and roles (queries)
–simulation of system behavior under the presence
of dysfunctional components
–verification of correctness (dynamic properties)
Limitations
• Model is qualitative
• Data entry is manual (no NLP)
• Learning curve for using the framework
to model a new biological domain is steep
• Definition of new queries for an existing
system requires use of 1st order logics
Thanks!
[email protected]
http://smi.stanford.edu/people/peleg