from domain ontologies to modeling ontologies to executable simulation models gregory a. silver...
TRANSCRIPT
From Domain Ontologies to Modeling Ontologies to Executable Simulation Models
Gregory A. Silver
Osama M. Al-Haj Hassan
John A. Miller
University of Georgia
2007 Winter Simulation Conference
OutlineI. Ontology Driven Simulation (ODS)
– Definition & Motivation– Historical Perspective
II. Web Based Resources for Modeling & Simulation– Domain Ontologies– Modeling Ontologies– Structured (e.g. databases) and Unstructured (e.g. papers) Sources
III. Development of an ODS Prototype– ODS Architecture– Ontology Mapping Tool & Markup Language Generation– Executable Model Generation
IV. ODS in Action: Two Examples– Hospital Emergency Department– Glycan Biosynthesis
I. Ontology Driven Simulation
• Definitions:– Domain Ontology – Knowledge in particular
domains is captured through defining concepts, their relationships, and relevant constraints.
• OWL (Web Ontology Language) is widely used for the Semantic Web.
– Ontology Driven Simulation – Simulation model development assisted/driven by application domain knowledge stored in ontologies.
• Motivation: Use the knowledge and data resident in domain ontologies to bootstrap the creation of simulation models.
Historical Perspective
• A Port Ontology for Automated Model Composition (Laing and Paredis 2003)
• Discrete-event Modeling Ontology (DeMO) (Miller, et al. 2004)
• Synthetic Environment Data Representation Ontology (sedOnto) (Bhatt, et al. 2005)
• Evaluation of the C2IEDM as an Interoperability Enabling Ontology (Turnitsa and Tolk 2005)
• Ontology Driven Framework for Simulation Modeling (Benjamin et al. 2005)
• Process Interaction Modeling Ontology for Discrete Event Simulation (PIMODES) (Lacy 2006)
II. Web Based Resources for Modeling & Simulation
– Creation of simulation models requires gathering of substantial amounts of knowledge and data.
– Sources of Information• Domain Ontologies – Domain Expertise
– GlycO – Glycomics Ontology
– EnzyO – Enzyme Ontology
– PMRO – Problem-oriented Medical Records Ontology
– Modeling Ontologies – Expertise in Modeling Techniques
• Discrete-event Modeling Ontology (DeMO)
– Online Databases• RK-Savio, BRENDA, KEGG
– Text Mining• PubMed
DeMO Top Level Classes
III. Development of an ODS PrototypeA. Goals
1. Support the use of Multiple Modeling Technologies
2. Tools for extracting and mapping Domain Ontologies
3. Support code generation for several simulation engines
B. The ODS Approach1. Discovery Phase – Search and Browse Multiple Ontologies
a. Relevant Domain Knowledge
b. Applicable Modeling Techniques
2. Mapping Phase – a. Connect and transform classes, properties and instances in Domain Ontologies to
those in Modeling Ontologies
b. Generate any additional instances required in Modeling Ontology
3. Code Generation Phasea. Two-Stage: OWL XML Code
• Advantage: Many simulation work off of an XML dialect such as the Petri Net Markup Language (PNML)
b. One-Stage: OWL Code• XML by itself is weak at expressing named relationships and constraints – so there is
the potential for information loss.
PMROOntology
GlycOOntology
ODS DesignTool Suite
DeMOOntology
DeMOInstances
Model MarkupGenerator
XPIM ePNMLBipartite Graph
LayoutAlgorithm
Graph LayoutAlgorithm
JSIMCode Generator
ARENACode Generator
ExecutableJSIM Model
ExecutableARENA Model
JSIM Petri NetEngine
JSIMExecution
Engine
ARENAExecution
Engine
ERPatientFlow
Modeling
GlycanBiosynthetic
Pathway Modeling
Ontology Editor, Browser,Viewer, and Alignment
Tools
SPARQL + XSLT
Event Oriented, StateOriented, Process Oriented,
Activity Oriented
AGLO
JavaProgram
JSIM Petri NetCode Generator
Executable PetriNet Model
Ontology Driven Simulation Architecture
Map PMRO classes to DeMO Classes DeMO Represention of Model (OWL Instances)
Generate Markup Language Instances
Ontology Mapping Tool & Markup Language Generation
<activity activityid="ClinicalExamination" activitytype="Facility" caption="Examination" "> <location x="331" y="314" /> <costdist distributiontype="Uniform" alpha="100.0"
beta="300.0" stream="0" /> <servicedist distributiontype="Uniform" alpha="300.0" beta="200.0" stream="0" /></activity>
XPIML Representation of Model
Executable Model Generation
<activity activityid="ClinicalExamination" activitytype="Facility" caption="Examination" "> <location x="331" y="314" /> <costdist distributiontype="Uniform" alpha="100.0"
beta="300.0" stream="0" /> <servicedist distributiontype="Uniform" alpha="300.0" beta="200.0" stream="0" /></activity>
Executable Model Generator
XPIML Representation of Model
JSIM Execution
IV. ODS in Action: Two Examples• Hospital Emergency Room
– PMRO JSIM– Process Interaction
• Glycan Biosynthesis– GlycO, EnzyO HFPN– Petri Nets
Hospital Emergency Room Example
PMROOntology
ODS Design ToolSuite
DeMOOntology
DeMOPIModel
Instances
Model MarkupGenerator
XPIMLInstances
JSIM Petri NetEngine
Knowledge Extraction
Model Construction
OWL Instance XPIML Instance
JSIM Specification
JSIM Execution
Biochemical Pathway ODS
GlycOOntology
EnzyOOntology
SystemsBiology
Ontology
ScientificLiterature
Biochemical- Pathways- ReactionKenitics
RK - SayioDatabase
ReactionLayers &
Constants
ODS Design ToolSuite
DeMOOntology
DeMOHFPNModel
Instances
Model MarkupGenerator
ePNMLInstances
JSIM Petri NetEngine
Knowledge Extraction
Model Construction
Biochemical Pathway for Glycan Biosynthesis
Michaelis-Menten Reaction Kinetics
v0 = Vmax[S]Km+[S]
Hybrid Functional Petri Nets
S1
E1
P1
E2
P2
R1 R2
ES EB RA
Glycan
[S1]
ES EB RA
RNA Protien Enzyme
[E1]
ES EB RA
Glycan
[P1]
ES EB RA
[E2]
ES EB RA
Glycan
[P2]
RNA Protien Enzyme
Substrate
Enzyme
Product
Enzyme
Product