Exploiting the Biophysical Semantics of Biosimulation Models

1
Exploiting the Biophysical Semantics of
Biosimulation Models
Daniel L. Cook 1, 2 John H. Gennari 3 Maxwell L.
Neal 3 Michal Galdzicki 3
1Physiology Biophysics, 2Biological
Structure 3Biomedical and Health
Informatics University of Washington, Seattle
2
The grand challenge
Organize the world's information and make it
universally accessible and useful.
3
Our (lesser) grand challenge
Organize the world's biosimulation models and
make them universally accessible and useful as
modules for testing scientific hypotheses and
predicting patient-specific health outcomes.
4
Requirements for success

• Modular plug-n-play
• extract and merge modules for specific needs
• Comprehensive
• multiscale and multidomain
• Operate with models in multiple languages
• CellML, SBML, JSim, MATLAB,...
• User-friendly software support
• shield users from mechanics of ontologies and
  queries

5
Multiscale
Multiple structural scales
12 organ systems
63 organ types
2 bodies
gtgt 100,000 molecule types
gt400 cell-part types
gt 100 elements
gt600 cell types
meters
1
10-12
extended from Hunter, P. J. Borg, T. K. (2003).
Nat Rev Mol Cell Biol 24(6)667-72.
6
Multiscale / Multidomain
Multiple structural scales
Multiple process domains
7
Outline

• Problem and current practice
• Biosimulation model-building
• MIRIAM annotation for molecular models
• An ontology-based solution
• Use-case examples of Semantic Simulation models
• Knowledge foundations for SemSim
• Foundational Model of Anatomy multiscale
  structure
• Ontology of Physics for Biology multidomain
  physics
• SemSim semantic architecture
• Physical model
• Computational model
• Summary of progress

8
Biosimulation modeling traditional practice
9
Current practice multiple languages
Declarative MML(JSim) SBML CellML BioPAX Proc
edural MATLAB Fortran
Alves, R., F. Antunes, et al. (2006). Tools for
kinetic modeling of biochemical networks Nat
Biotechnol 24(6) 667-72.
10
Current practice multiple platforms



multiscale/multidomain capabilities
Alves, R., F. Antunes, et al. (2006). Tools for
kinetic modeling of biochemical networks Nat
Biotechnol 24(6) 667-72.
11
Current practice multiple archives

• CellML (100s)
• biochemical
• cardiovascular
• mass transport

• BioModels (100s)
• biochemical
• cell biology

• NSR Physiome (100s)
• biochemical
• cardiovascular
• mass transport
• respiratory
• ...

biosim model code

• Other (n ?)
• biochemical
• cardiovascular
• respiratory

• MATLAB (n ?)
• biochemical
• cardiovascular
• electrophysiology

12
Challenge search for relevant models

• CellML (100s)
• biochemical
• cardiovascular
• mass transport

• BioModels (100s)
• biochemical
• cell biology

biosim model code
?
biosim model code
?

• NSR Physiome (100s)
• biochemical
• cardiovascular
• mass transport
• respiratory
• ...

biosim model code
?
biosim model code
biosim model code
?
?

• Other (n ?)
• biochemical
• cardiovascular
• respiratory

• MATLAB (n ?)
• biochemical
• cardiovascular
• electrophysiology

13
Challenge extract and merge modules

• CellML (100s)
• biochemical
• cardiovascular
• mass transport

• BioModels (100s)
• biochemical
• cell biology

biosim model code
?
biosim model code
?

• NSR Physiome (100s)
• biochemical
• cardiovascular
• mass transport
• respiratory
• ...

?
biosim model code
?
?
biosim model code
biosim model code
?

• Other (n ?)
• biochemical
• cardiovascular
• respiratory

• MATLAB (n ?)
• biochemical
• cardiovascular
• electrophysiology

14
Annotation helps MIRIAM for systems biology
pointer or citation to model reference
description
MIRIAM constituent
biosim model code
Le Novere, N., A. Finney, et al. (2005). Minimum
information requested in the annotation of
biochemical models (MIRIAM) Nat Biotechnol
23(12) 1509-15.
15
Annotate structures
specify physical reactants and compartments
MIRIAM constituent
biosim model code
Le Novere, N., A. Finney, et al. (2005). Minimum
information requested in the annotation of
biochemical models (MIRIAM) Nat Biotechnol
23(12) 1509-15.
16
Annotate physical laws
MIRIAM constituent
biosim model code
specify reaction kinds and rate laws
Le Novere, N., A. Finney, et al. (2005). Minimum
information requested in the annotation of
biochemical models (MIRIAM) Nat Biotechnol
23(12) 1509-15.
17
Recently MIRIAM URI resources
URI pointers to on-line knowledge resources
MIRIAM resources
GO
ChEBI
biosim model code
SBO
Laibe, C. and N. Le Novere (2007). MIRIAM
Resources tools to generate and resolve robust
cross-references in Systems Biology BMC Syst Biol
1 58.
18
MIRIAM limited scope
Multiple structural scales
GO
ChEBI
Multiple process domains
SBO
19
The SemSim proposal
Multiple structural scales
Develop machine-readable semantic methods for
annotating, analyzing and linking models across
all biological stuctural scales and process
domains.
GO
ChEBI
Multiple process domains
SBO
20
Outline

• Problem and current practice
• Biosimulation model-building
• MIRIAM annotation for molecular models
• An ontology-based solution
• Use-case examples of Semantic Simulation models
• Knowledge foundations for SemSim
• Foundational Model of Anatomy multiscale
  structure
• Ontology of Physics for Biology multidomain
  physics
• SemSim semantic architecture
• Physical model
• Computational model
• Summary of progress

21
Biosimulation modeling
Execute model code
Results
Data
Encode in a computational language
Ensure unit consistency
Generate physics-based equations
Specify physical theory for the system
Specify biological entities in the system
?
Formulate question
22
SemSim modeling software
Execute model code
Results
Data
Encode in a computational language
Ensure unit consistency
Generate physics-based equations
CodeGen
Specify physical theory for the system
Specify biological entities in the system
SemGen
?
Formulate question
23
SemSim use-case Cardiovascular integration
Vascular smooth muscle (D. Cook, B. Carlson)
Baroreceptor reflex (D. Beard)
Cardiovascular system (M.Neal, JBB)
Gennari, J. H., M. L. Neal, B. E, Carlson, D. L.
Cook (2008) Integration of multi-scale
biosimulation models via light-weight
semantics Pac Symp Biocomput (414-425)
24
SemSim use-case Cardiovascular integration
Vascular smooth muscle (D. Cook, B. Carlson)
What happens to heart rate and blood pressure if
vascular smooth muscle Ca uptake increases?
Baroreceptor reflex (D. Beard)
Cardiovascular system (M.Neal, JBB)
Gennari, J. H., M. L. Neal, B. E, Carlson, D. L.
Cook (2008) Integration of multi-scale
biosimulation models via light-weight
semantics Pac Symp Biocomput (414-425)
25
SemSim use-case 1 create SemSim models
1. create SimSim models
Gennari, J. H., M. L. Neal, B. E, Carlson, D. L.
Cook (2008) Integration of multi-scale
biosimulation models via light-weight
semantics Pac Symp Biocomput (414-425)
26
SemSim use-case 1 create SemSim models
Semantic Simulation model a light-weight Protégé
ontology represents the biological meaning and
mathematical structure of a biosimulation module
Gennari, J. H., M. L. Neal, B. E, Carlson, D. L.
Cook (2008) Integration of multi-scale
biosimulation models via light-weight
semantics Pac Symp Biocomput (414-425)
27
SemSim use-case 1 merge SemSim models
2. merge ontology modules using Prompt plug-in to
Protege
Gennari, J. H., M. L. Neal, B. E, Carlson, D. L.
Cook (2008) Integration of multi-scale
biosimulation models via light-weight
semantics Pac Symp Biocomput (414-425)
28
SemSim use-case 1 encode as JSim-MML
VSM
SemSim
CV
BARO
SemSim
SemSim
CV
SemSim
3. encode SemSim model as JSim-MML
Gennari, J. H., M. L. Neal, B. E, Carlson, D. L.
Cook (2008) Integration of multi-scale
biosimulation models via light-weight
semantics Pac Symp Biocomput (414-425)
29
SemSim use-case 2 enhance arterial dynamics
How does fluid inertia affect the aortic pulse
pressure profile?
T. Arts (U. Masstricht)
CV
CV-CA
SemSim
CA sys art
SemSim
Neal, M. L., J. H. Gennari, T. Arts, D. L.
Cook Advances in semantic representation of
multiscale biosimulations a case study in
merging models Pac Symp Biocomput (submitted)
30
SemSim use-case 2 procedural declarative
CV
CV-CA
SemSim
CA sys art
SemSim
Declarative
Procedural
Neal, M. L., J. H. Gennari, T. Arts, D. L.
Cook Advances in semantic representation of
multiscale biosimulations a case study in
merging models Pac Symp Biocomput (submitted)
31
SemSim use-case 2 procedural declarative

• General issues
• time-scale separation
• flattening complex data code structures
• tracing variable dependencies

CV
CV-CA
SemSim
CA sys art
SemSim
Declarative
Procedural
Neal, M. L., J. H. Gennari, T. Arts, D. L.
Cook Advances in semantic representation of
multiscale biosimulations a case study in
merging models Pac Symp Biocomput (submitted)
32
Beginnings of a SemSim Archive
SemSim Archive
VSM
SemSim
CV
BARO
SemSim
SemSim
CV
CV-CA
SemSim
CA sys art
CA sys art
SemSim
SemSim
33
Outline

• Problem and current practice
• Biosimulation model-building
• MIRIAM annotation for molecular models
• An ontology-based solution
• Use-case examples of Semantic Simulation models
• Knowledge foundations for SemSim
• Foundational Model of Anatomy multiscale
  structure
• Ontology of Physics for Biology multidomain
  physics
• SemSim semantic architecture
• Physical model
• Computational model
• Summary of progress

34
Multiscale/multidomain knowledge foundations
1
Foundational Model of Anatomy multiscale
structure
structural knowledge
FMA
GO
ChEBI
biosim model code
SemSim model
SBO
OPB
2
physics-based process knowledge
Ontology of Physics for Biology multidomain
physics
35
Foundational Model of Anatomy (FMA)
anatomical types
Rosse, C. and J. V. Mejino Jr (2005).
Foundational Model of Anatomy. http//sig.biostr.w
ashington.edu/projects/fm/AboutFM.html
a Protégé or OWL ontology
36
Foundational Model of Anatomy (FMA)
partonomy
Rosse, C. and J. V. Mejino Jr (2005).
Foundational Model of Anatomy. http//sig.biostr.w
ashington.edu/projects/fm/AboutFM.html
37
Multiscale/multidomain foundations for SemSim
1
Foundational Model of Anatomy multiscale
structure
structural knowledge
FMA
GO
ChEBI
biosim model code
SemSim model
SBO
OPB
2
physics-based process knowledge
Ontology of Physics for Biology multidomain
physics
Cook, D. L., J. L. V. Mejino, M. L. Neal, J. H.
Gennari (2008) Bridging Biological Ontologies
and Biosimulation The Ontology of Physics for
Biology Proceedings of the AMIA Fall Symposium
(in press)
a Protégé ontology
38
OPB simple representational schema
e.g., heart, blood in aorta, protein kinase A,
Ca
e.g., mass, pressure, molar flow, concentration
e.g., Ohms law, Law of mass action,
Conservation of mass
Cook, D. L., J. L. V. Mejino, M. L. Neal, J. H.
Gennari (2008) Bridging Biological Ontologies
and Biosimulation The Ontology of Physics for
Biology Proceedings of the AMIA Fall Symposium
(in press)
39
OPB schema maps to biosimulation code
Physical entities known at best by annotation or
variable names
real Paorta(t)   mmHg // Aorta
pressure real PSysVein(t)   mmHg   //
Systemic vein pressure real FSysArt(t)
ml/sec // Systemic artery flow real Rartcap
0.7 mmHgsec/ml  // Arterial
resistance FSysArt (Paorta - PSysVein) /
Rartcap // Ohm's Law
Gennari, J. H., M. L. Neal, B. E, Carlson, D. L.
Cook (2008) Integration of multi-scale
biosimulation models via light-weight
semantics Pac Symp Biocomput (414-425)
40
OPB foundational theory system dynamics

• Engineering system dynamics
• Bond graph theory
• Karnopp, Margolis, Rosenberg (1968)
• PHYSYS - Physical Systems Ontology
• Borst, Top, Akkermans (1994)
• Biochemical system dynamics
• Network thermodynamics
• Oster, Perelson, Katchalsky (1973)
• Mickulecky (1983)
• Beard, Qian (2008)

41
OPB system dynamics Kinetic properties
42
OPB encodes Temporal integral dependency
43
OPB encodes Constitutive dependencies
44
Schema maps to OPB classes
Ontology of Physics for Biology
Cook, D. L., J. L. V. Mejino, M. L. Neal, J. H.
Gennari (2008) Bridging Biological Ontologies
and Biosimulation The Ontology of Physics for
Biology Proceedings of the AMIA Fall Symposium
(in press)
45
OPB multiple Physics domains
Kinetic domains in which biological processes
occur
46
OPB Physical entities
Entities as in FMA, GO, ChEBI, etc.
47
OPB Physical properties
Generalized kinetic properties of system dynamics
48
OPB Kinetic properties for each Physical domain
A Kinetic property for each Physics domain
49
OPB encodes Physical dependencies
Empirically-determined laws by which Kinetic
properties depend upon one another
50
OPB encodes Constitutive dependencies
Dependencies for each Kinetic domain
51
Outline

• Problem and current practice
• Biosimulation model-building
• MIRIAM annotation for molecular models
• An ontology-based solution
• Use-case examples of Semantic Simulation models
• Knowledge foundations for SemSim
• Foundational Model of Anatomy multiscale
  structure
• Ontology of Physics for Biology multidomain
  physics
• SemSim semantic architecture
• Physical model
• Computational model
• Summary of progress

52
SemSim Physical model Computational model
structural knowledge
SemSim model
FMA
Physical model
Computational model
biosim model
GO
ChEBI
simulation code
variable
SBO
OPB
equation
physics-based process knowledge
53
SemSim Physical model
structural knowledge
SemSim model
FMA
Physical model
Computational model
biosim model
GO
Physical entity
ChEBI
simulation code
has_property
Physical property
variable
SBO
has_player
OPB
equation
physics-based process knowledge
Physical model (the OPB schema) is a declarative
representation of the structure and physics of
the modeled biological system
54
SemSim Computational model
structural knowledge
SemSim model
FMA
Physical model
Computational model
biosim model
GO
Model source
Physical entity
ChEBI
simulation code
has_property
has_data
Physical property
variable
SBO
use / return
has_player
OPB
equation
physics-based process knowledge
Computational model represents the source
metadata and mathematical structure of the
biosimulation model
55
SemGen semiautomated SemSim builder
structural knowledge
SemSim model
FMA
Physical model
Computational model
biosim model
GO
Model source
Physical entity
ChEBI
simulation code
has_property
has_data
Physical property
variable
SBO
use / return
has_player
OPB
equation
physics-based process knowledge
SemGen
in development, M. Neal
56
SemGen Step 1 model source metadata
structural knowledge
SemSim model
FMA
Physical model
Computational model
biosim model
GO
Model source
Physical entity
ChEBI
simulation code
has_property
has_data
Physical property
variable
SBO
use / return
has_player
OPB
equation
physics-based process knowledge
SemGen
57
SemGen Step 2 map model variables
structural knowledge
SemSim model
FMA
Physical model
Computational model
biosim model
GO
Model source
Physical entity
ChEBI
simulation code
has_property
has_data
Physical property
variable
SBO
use / return
has_player
OPB
equation
physics-based process knowledge
SemGen
58
SemGen Step 3 map Properties to Physical
entities
structural knowledge
SemSim model
FMA
Physical model
Computational model
biosim model
GO
Model source
Physical entity
ChEBI
simulation code
has_property
has_data
Physical property
variable
SBO
use / return
has_player
OPB
equation
physics-based process knowledge
SemGen
59
SemGen Step 4 map equations
structural knowledge
SemSim model
FMA
Physical model
Computational model
biosim model
GO
Model source
Physical entity
ChEBI
simulation code
has_property
has_data
Physical property
variable
SBO
use / return
has_player
OPB
equation
physics-based process knowledge
SemGen
60
SemSim model taxonomies
Physical model
Computational model
61
Building and using a SemSim Archive
CellML
biosim model
BioModels
biosim model
SemSim model
patient specific model
SemSim model
NSR Physiome
biosim model
SemSim model
SemSim model
SemSim
SemSim model
biosim model
MATLAB
PSM
biosim model
Other
62
Outline

• Problem and current practice
• Biosimulation model-building
• MIRIAM annotation for molecular models
• An ontology-based solution
• Use-case examples of Semantic Simulation models
• Knowledge foundations for SemSim
• Foundational Model of Anatomy multiscale
  structure
• Ontology of Physics for Biology multidomain
  physics
• SemSim semantic architecture
• Physical model
• Computational model
• Summary of progress

63
Elements of multiscale/multidomain SemSim
Foundational Model of Anatomy multiscale
structure
structural knowledge
SemSim model multiscale/multidomain declarative
annotation
FMA
GO
ChEBI
biosim model code
SemSim model
SBO
CodeGen
SemGen
OPB
physics-based process knowledge
Ontology of Physics for Biology multidomain
system dynamics
64
Progress toward success

• Modular plug-n-play
• 2 successful SemSim use-case demos
• Comprehensive
• FMA and OPB as knowledge foundations
• Operate with models in multiple languages
• SemSim is language-independent
• User-friendly software support
• SemGen and CodeGen in developmennt

65
Acknowledements

• UW Semantic simulation team
• Maxwell L. Neal (Grad student)
• Michal Galdzicki (Grad student)
• John H. Gennari, PhD (Assoc Prof)
• Daniel L. Cook, MD, PhD (Res Prof)

• UW collaborators
• Cornelius Rosse
• Onard Mejino
• James Brinkley
• James B. Bassingthwaighte
• Erik Butterworth
• Herbert Sauro
• Hong Qian
• Adriana Emmi
• Fred Bookstein

• Other consultants
• Theo Arts
• Daniel Beard
• Peter Hunter
• Andrew McCulloch
• Frank Sachse

Partial funding from NIH MLN, MG T15
LM007442-06 DLC, JHG R01HL087706-01
66
questions?
SemSim
67
SemSim model frames and links
Physical model
Computational model
68
OPB knowledge Dependency Property links
69
OPB template knowledge for simulation code
F12 ( P1 - P2 ) / R12
70
OPB law of mass action is a control dependency
Flow
rate law
Force
Displacement
Force
71
OPB network thermodynamics Flow f(A)
JRP
J kRR - kPP
R, P
ARP - µP µR
R molar gas constant T temperature µP
chemical potential
from Tuszynski Kurzynski (2003)
72
OPB Transformer dependency
73
OPB Control dependency
74
OPB thermodynamics of constitutive dependencies
heat energy
potential energy
kinetic energy
75
OPB thermodynamic conservation of energy
constraint
heat energy
potential energy
kinetic energy
?
conservation of energy