From Domain Ontologies to Modeling Ontologies to Executable

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

Outline I. Ontology Driven Simulation (ODS) – – II. Definition & Motivation Historical Perspective Web Based Resources for Modeling & Simulation – – – III. Domain Ontologies Modeling Ontologies Structured (e. g. databases) and Unstructured (e. g. papers) Sources Development of an ODS Prototype – – – IV. ODS Architecture Ontology Mapping Tool & Markup Language Generation Executable Model Generation 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 (De. MO) (Miller, et al. 2004) • Synthetic Environment Data Representation Ontology (sed. Onto) (Bhatt, et al. 2005) • Evaluation of the C 2 IEDM 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 – Glyc. O – Glycomics Ontology – Enzy. O – Enzyme Ontology – PMRO – Problem-oriented Medical Records Ontology – Modeling Ontologies – Expertise in Modeling Techniques • Discrete-event Modeling Ontology (De. MO) – Online Databases • RK-Savio, BRENDA, KEGG – Text Mining • Pub. Med

De. MO Top Level Classes

III. Development of an ODS Prototype A. Goals 1. 2. 3. B. Support the use of Multiple Modeling Technologies Tools for extracting and mapping Domain Ontologies Support code generation for several simulation engines The ODS Approach 1. Discovery Phase – Search and Browse Multiple Ontologies a. b. 2. Mapping Phase – a. b. 3. Relevant Domain Knowledge Applicable Modeling Techniques Connect and transform classes, properties and instances in Domain Ontologies to those in Modeling Ontologies Generate any additional instances required in Modeling Ontology Code Generation Phase a. Two-Stage: OWL XML Code • b. Advantage: Many simulation work off of an XML dialect such as the Petri Net Markup Language (PNML) One-Stage: OWL Code • XML by itself is weak at expressing named relationships and constraints – so there is the potential for information loss.

Ontology Driven Simulation Architecture

Ontology Mapping Tool & Markup Language Generation Map PMRO classes to De. MO Classes De. MO Represention of Model (OWL Instances) Generate Markup Language Instances XPIML Representation of Model <activityid="Clinical. Examination" activitytype="Facility" caption="Examination" "> <location x="331" y="314" /> <costdistributiontype="Uniform" alpha="100. 0" beta="300. 0" stream="0" /> <servicedistributiontype="Uniform" alpha="300. 0" beta="200. 0" stream="0" /> </activity>

Executable Model Generation XPIML Representation of Model <activityid="Clinical. Examination" activitytype="Facility" caption="Examination" "> <location x="331" y="314" /> <costdistributiontype="Uniform" alpha="100. 0" beta="300. 0" stream="0" /> <servicedistributiontype="Uniform" alpha="300. 0" beta="200. 0" stream="0" /> </activity> JSIM Execution Executable Model Generator

IV. ODS in Action: Two Examples • Hospital Emergency Room – PMRO JSIM – Process Interaction • Glycan Biosynthesis – Glyc. O, Enzy. O HFPN – Petri Nets

Hospital Emergency Room Example Knowledge Extraction Model Construction

OWL Instance JSIM Specification JSIM Execution XPIML Instance

Biochemical Pathway ODS Knowledge Extraction Model Construction

Biochemical Pathway for Glycan Biosynthesis Michaelis-Menten Reaction Kinetics v 0 = Vmax[S] Km+[S] Hybrid Functional Petri Nets S 1 P 1 Substrate R 1 Product Enzyme [E 1] ES EB RA Glycan R 2 Enzyme [S 1] P 2 [P 1] ES EB RA RNA Protien Enzyme [P 2] [E 2] ES EB RA Glycan RNA ES EB RA Protien Enzyme Glycan