Bridging Biological Ontologies and Biosimulation: The Ontology of Physics for Biology

Bridging Biological Ontologies and Biosimulation:The Ontology of Physics for Biology

Daniel L. Cook 1, 2

John H. Gennari 3

Jose L. V. Mejino 2

Maxwell L. Neal 3

1Physiology & Biophysics, 2Biological Structure3Biomedical and Health InformaticsUniversity of Washington, Seattle

AMIA 2008, Washington, DC

63 organ types

2 bodies

>> 100,000 molecule types

>400 cell-part types

12 organ systems

> 100 elements

>600 cell types

Available bioinformatics for “multiscale” structure

extended from Hunter, P. J. & Borg, T. K. (2003). Nat Rev Mol Cell Biol 24(6):667-72.

Foundational Model of Anatomy

Gene Ontology

ChEBI

Cell Type

63 organ types

2 bodies



12 organ systems

> 100 elements

>600 cell types

No bioinformatics for multidomain processes

fluidssolids

diffusionheat transfer

myocardial contraction, leg motion…

electrochemistry transmembrane potential, action potential…chemical kinetics metabolism, gene expression, cell signaling…

body temperature regulation…intracellular calcium dynamics…

blood flow, respiratory gas flow…

Physical Physical domainsdomains

Domain Process

63 organ types

2 bodies



12 organ systems

> 100 elements

>600 cell types

Bioinformatic problem: query process knowledge

fluidssolids







Domain Process

• How is blood pressure controlled?• Which nerves control blood

pressure?

63 organ types

2 bodies



12 organ systems

> 100 elements

>600 cell types

Processes encoded as biosimulations models

fluidssolids







Domain Process

physics-basedbiosimulation model

63 organ types

2 bodies



12 organ systems

> 100 elements

>600 cell types

Available models constitute “physiome”

fluidssolids







Domain Process

PhysiomePhysiome

Hunter, P. J. & Borg, T. K. (2003). Nat Rev Mol Cell Biol 24(6):667-72.

63 organ types

2 bodies



12 organ systems

> 100 elements

>600 cell types

Physiome problem: reuse and merge models

fluidssolids







Domain Process

PhysiomePhysiome

Hunter, P. J. & Borg, T. K. (2003). Nat Rev Mol Cell Biol 24(6):667-72.

physics-basedbiosimulation model

Proposal for a solution:

Semantics of biosimulation models can be encoded as ontologies and

mapped to reference ontologies.

Biosimulation

model codeSemSimReference

ontologies

Outline:

Biosimulation

model codeSemSimOPB, FMA,

GO, CheBI, etc.

• Problems: biosimulation, bioinformatics

• SemSim ontology• Ontology of Physics for Biology

(OPB)• Conclusion



structuralknowledge

physicsknowledge

physics-basedprocess

biosimulation

fluidssolids

chemical kinelectrochem


Time

In practice: code is hand-crafted

Biophysicists and bioengineers encode physics-based

mathematical models of biological processes

In practice: code is formal — meaning is implicit

structuralknowledge

physicsknowledge

fluidssolids



real Paorta(t) mmHg; // Pressure of aortareal PSysVein(t) mmHg; // Pressure of systemic veinreal FSysArt(t) ml/sec; // Flow in systemic arteryreal Rartcap = 0.7 mmHg*sec/ml; // Arterial resistance

FSysArt = (Paorta - PSysVein) / Rartcap; // Ohm's Law

physiological variable

names are arbitrary

anatomical participants

known only by annotation

variable dependencies known only by

annotation

In practice: multiple, incompatible languages

structuralknowledge

physicsknowledge

fluidssolids



JSim, SBML, CellML, MatLab,

others…


biosimulation

In practice: 100’s of models in linguistic silos

structuralknowledge CellML

SBML

JSim

MatLab

other

physicsknowledge

fluidssolids




biosimulation


biosimulation


biosimulation


biosimulation


biosimulation

Opportunity: a reservoir of process knowledge

CellML

SBML

JSim

MatLab

other

Problem: barriers to biosimulation model reuse

CellML

SBML

JSim

MatLab


biosimulation

other

JSim

?

??

How to find, merge and re-encode models?

Problem: no access for bioinformatic queries

CellML

SBML

JSim

MatLab

other

Q & A

How to query knowledge of biological processes? SparQL

Two fields, two problems:

• Find models of blood pressure control.

• Which models include neural-control?

Biosimulation — re-use biosimulation models

• How is blood pressure controlled?• Which nerves control blood pressure?

Bioinformatics — query process knowledge

Outline:

Biosimulation


GO, CheBI, etc.



(OPB)• Conclusion



Solution: encode SemSim ontological maps…

CellML

SBML

JSim

MatLab

other

SemSimSemSimSemSimSemSimSemSimOWL

SemSimsemantic maps of

biosimulation models

OPB, FMA, GO, CheBI, etc.

…and annotate to reference ontologies

CellML

SBML

JSim

MatLab

other

SemSimSemSimSemSimSemSimSemSimOWL

annotate SemSim components to

reference ontologies

SemSimsemantic maps of


: Paorta PSysVein FSysArt Rartcap:: FSysArt =….:

structuralknowledge

physicsknowledge

fluidssolids



SemSim — biosimulation ontological map

SemSim model

biosimulation code

Computational model

Physicalmodel

Gennari, J. H., M. L. Neal, B. E, Carlson, D. L. Cook (2008) Integration of multi-scale biosimulation models via light-weight semanticsPac Symp Biocomput (414-425)

structuralknowledge

physicsknowledge

fluidssolids




SemSim — step 1: represent math structure

SemSim modelComputational

model

Data structure

biosimulation code

Physicalmodel

represent variable as individuals of class

Data structure

structuralknowledge

physicsknowledge

fluidssolids




SemSim — step 1: represent math structure

SemSim modelComputational

model

Computation

Data structure

use / return

biosimulation code

Physicalmodel

represent variable as individuals of class

Data structure

represent equations as individuals of class

Computation


SemSim — step 2: represent biological meaning

SemSim model

Physicalproperty

Physicalmodel

Computational model

Computation

Data structure

use / return

biosimulation code

structuralknowledge

physicsknowledge

fluidssolids



e.g., volume, pressure, molar flow, chemical

amount



SemSim model

Physicalentity

has_property

Physicalproperty

Physicalmodel

Computational model

Computation

Data structure

use / return

biosimulation code

structuralknowledge

physicsknowledge

fluidssolids



e.g., heart, blood in aorta, protein

kinase, folate, Ca++


amount



SemSim model

Physicalentity

has_property

Physical dependency

Physicalproperty

Physicalmodel

Computational model

Computation

Data structure

use / return

biosimulation code

has_player

structuralknowledge

physicsknowledge

fluidssolids



e.g., heart, blood in aorta, protein

kinase, folate, Ca++


amount

e.g., Ohm’s law, law of mass action, mass conservation


structuralknowledge

physicsknowledge

fluidssolids



SemSim model

Physicalentity

has_property

Physical dependency

Physicalproperty

has_player

Physicalmodel

Computational model

Computation

Data structure

use / return

biosimulation codeGO

ChEBI

FMA

Map to reference ontologies of structure


structuralknowledge

physicsknowledge

fluidssolids



SemSim model

Physicalentity

has_property

Physical dependency

Physicalproperty

has_player

Physicalmodel

Computational model

Computation

Data structure

use / return


ChEBI

FMA

OPB

Map to reference ontology of physics — OPB

Outline:

Biosimulation


GO, CheBI, etc.



(OPB)• Conclusion



OPB foundational theory — system dynamics

Engineering system dynamics• Bond graph theory

Karnopp, Margolis, Rosenberg (1968)• EngMath - Ontology for Engineering

MathematicsGruber, Olsen (1994)

• PHYSYS - Physical Systems OntologyBorst, Top, Akkermans (1994)

Biochemical system dynamics• Network thermodynamics

Oster, Perelson, Katchalsky (1971)Mickulecky (1983)Beard, Qian (2008)

OPB representational goals

• Represent abstractions used in physics-based biosimulations—not a theory of “reality”.

• Adhere to OBO principles.

• Implement in OWL; deploy to OBO and BioPortal.

OPB:Physics analytical entity

OPB

A Physics analytical entity is an abstraction of the real world created

within the science of classical physics for the description of physical entities

and the analysis of physical processes.

OPB:Physical entity

OPB




A Physical entity is a spatial, temporal, or energetic abstraction

of the physical world.

OPB:Physical property

OPB




A Physical property is a quantifiable attribute of a physical entity whose

value can be determined by physical measurement at a moment in time.



chemical kinetics

volume flow pressure volume pressure momentum

velocity force displacement solid momentum

molar flow chemical potential chemical amount ----

ionic current voltage charge ----

particle flow chemical potential particle number ----

heat flow temperature heat amount ----

fluids

solids

electrophysiology

diffusion

heat transfer

Physical property organizing principle

Physical domain

chemical kinetics

volume flow pressure volume pressure momentum

velocity force displacement solid momentum

molar flow chemical potential chemical amount ----

ionic current voltage charge ----

particle flow chemical potential particle number ----

heat flow temperature heat amount ----

fluids

solids

electrophysiology

diffusion

heat transfer

Physical property class hierarchy

Physical domain

OPB

A Flow subclass for each physical

domain

Physical property by domain

OPB




A Physical property is a quantifiable attribute of a physical entity whose

value can be determined by physical measurement at a moment in time.



A Physical dependency is a quantitative dependency between the magnitudes of two or more physical

properties according to a physical law.

Physical dependency

Physical dependency organizing principle

A Physical dependency is a quantitative dependency between the magnitudes of two or more physical

properties according to a physical law.

Axiomatic physical dependency

Constitutive physical dependency

Flow

e.g., “Ohm’s law”

Force


Flow

Force

Dis

plac

emen

t

Force

e.g., “Hooke’s law”


Flow

Force

Dis

plac

emen

t

Force

Mom

entu

m

Flow

Physical dependency class hierarchy OPB

A Resistive dependency subclass for each physical domain

OPB

Physical dependency by domain

: Paorta PSysVein FSysArt Rartcap::: FSysArt =….:

OPB-SemSim working example

SemSim model

Physicalentity

has_property

Physical dependency

Physicalproperty

has_player

Physicalmodel

Computational model

Computation

Data structure

use / return

model code

Neal, M. L., J. H. Gennari, T. Arts, D. L. Cook (2009)Advances in semantic representation of multiscale biosimulations: A case study in merging modelsPac Symp Biocomput (in press)

Conclusion

CellML

SBML

JSim

MatLab

other

SemSimSemSimSemSimSemSimSemSim

OPB FMA GO

ChEBI etc.

Acknowledgements

SemSim / OPB team• Maxwell L. Neal (Grad student)• Michal Galdzicki (Grad student)• John H. Gennari, PhD (Assoc Prof)• Daniel L. Cook, MD, PhD (Res

Prof)

Bioinformatics• Cornelius Rosse• Onard Mejino• James Brinkley• Todd Detwiler

Partial funding from NIH MLN, MG: T15 LM007442-06 DLC, JHG: R01HL087706-01

UW contributorsBiophysics / biosimulation• James B. Bassingthwaighte• Herbert Sauro• Erik Butterworth• Hong Qian• Adriana Emmi• Fred Bookstein


structuralknowledge

physicsknowledge

fluidssolids



Next steps…

SemSim model

Physicalentity

has_property

Physical dependency

Physicalproperty

has_player

Physicalmodel

Computational model

Computation

Data structure

use / return


ChEBI

FMA

SemGenparse code

access classeswrite new code

Ontology of

Physics for

Biology

VSM

JSim

BARO

JSim

CV

JSim

VSM

SemSim

BARO

SemSim

CV

SemSim

CV+

SemSim

CV+

JSim

SemSim use-case 1: reuse legacy models

Gennari, J. H., M. L. Neal, B. E, Carlson, D. L. Cook (2008) Integration of multi-scale biosimulation models via light-weight semanticsPac Symp Biocomput (414-425)

3. encode merged SemSimas JSim model

2. use Prompt plug-in to Protégé to analyze and merge SemSim models

1. create SemSim models of JSim


CV

JSim

VSM

SemSim

BARO

SemSim

VSM

JSim

BARO

JSim

CV+

SemSim

CV+

JSim

CircAdapt

MATLAB

CAsys art

JSim

CV

SemSim CV-CAsys art

JSim

Neal, M. L., J. H. Gennari, T. Arts, D. L. Cook (2009)Advances in semantic representation of multiscale biosimulations: A case study in merging modelsPac Symp Biocomput (in press)

CAsys art

SemSim

CV-CA

SemSim

CAsys art

SemSim use-case 2: reuse merged model

1. reuse archived SemSim model

2. create and merge model in different

language

FOL

SemSim

METH

SemSim

FOMC

SemSim

Fol-all

SemSim

Fol-all

JSim

SemSim use-case 3: folate chemical kinetics

Mike Galdzicki, J. H. Gennari, M. L. Neal, D. L. Cook (work in progress)

1. create SemSim models of folate metabolism from published descriptions

FOL

METH

FOMC

2. merge FOL & METH SemSim models

Nijhout, et al. (2004)

Reed, et al. (2004)

Reed, et al. (2006)

FOMC

JSim

3. encode SemSim models in JSim;

compare model output

Model are complex but can be parsed

Yngve, G., J. F. Brinkley, D. L. Cook, L. G. Shapiro (2007) A Model Browser for BiosimulationAMIA Annu Symp Proc (836-40)

real Paorta(t) mmHg; // Pressure of aortareal PSysVein(t) mmHg; // Pressure of systemic veinreal FSysArt(t) ml/sec; // Flow in systemic arteryreal Rartcap = 0.7 mmHg*sec/ml; // Arterial resistanceFSysArt = (Paorta - PSysVein) / Rartcap; // Ohm's Law

physical variables

physical participants

variable dependencies

Ontological maps can be queried for facts

Aortic blood pressure depends on vagus nerve firing rate.

aortic blood pressure

vagus nerve firing rate

Documents

Bridging Biological Ontologies and Biosimulation: The Ontology of Physics for Biology