About Ontology Development

From Nemo

Jump to: navigation, search

Contents

Project Summary

The goal of this project is to design and implement ERP ontologies that can be used for representation, integration, and meta-analyses of ERP experiment results.

Personnel

Point Person(s): Gwen Frishkoff, Allen Malony, Don Tucker, & Dejing Dou
NEMO Ontology Task Force: Gwen Frishkoff, Dejing Dou, Bob Frank, Paea LePendu, Jason Sydes, & Haishan Liu
NEMO EEG/MEG Consortium: John Connolly, Tim Curran, Kerry Kilborn, Dennis Molfese, Charles Perfetti
Outside Collaborators: Jessica Turner (fBIRN), Angela Laird (BrainMap), Neuroscience Information Framework (Maryanne Martone)

Project Description

The ontology development process can be broken down into several steps: (a) specification and documentation of domain concepts and concept definitions (b) logic-based formalization of concepts (i.e., implementation of the ontology in a computer language, such as OWL); and (c) ontology-based applications, e.g., meta-anaysis of datasets that have been annotated using EEG and ERP ontologies.

(a) Specification of ontology concepts. Domain experts (GF & soon the NEMO consortium) define each concept (class, relation) with respect to two "basic" or foundational ontologies, BFO (Pierre Grenon et al.) and OBO-Relations (Barry Smith et al.). Following OBO "best practices," we reuse concepts from related ontologies, such as BirnLex/ !NeuroLex whenever possible. See "References" below for documentation of these source ontologies.

(b) Formalization of ontology (Protege/OWL). We are currently working on a third version of NEMO_ERP owldoc. Our strategy is to translate the CMAPs that are developed initially by domain experts (with some knowledge of BFO and OBO-rel) into formal ontologies using Protege. HL is currently responsible for the OWL implementation.

(c) Application of ontology (labeling of data, meta-analysis using Matlab/Ontology Database tools). ERP ontologies will be used for annotation of single-subject ERP patterns. These annotated data will then be stored in our ontology database to enable cross-study, cross-lab meta-analyses. We currently have >500 single-subject datasets in our database, representing data from 19 language processing experiments. Our pilot meta-analysis is focused on classification and integration of frontal negativities peaking between 250 and 450 ms (MFN, fN4, and N3 labeled patterns). We are examining modulation of these patterns as a function of three cognitive subtractions: semantic priming (unrelated–related), lexical priming (pseudowords-words), and episodic memory for words (new – old).

Components

Classes
Domain-specific classes (such as 'ERP_Pattern' or 'Centroid') are defined with respect to the Basic Formal Ontology (BFO), which was developed at the Institute for Formal Ontology and Medical Information Science (Saarland University) by Barry Smith and associates. Building our class ontologies on top of BFO will allow us to more easily integrate with other BFO users, such as fBIRN. See http://www.ifomis.org/bfo for additional information.
Properties
Domain-specific relations (such as rostral_to, left_of, has_measurement) are defined in reference to (i.e., as sub-relations of) the basic relations found in OBO-Relations (http://www.obofoundry.org/ro/).

Note that OBO-R references BFO, as well. These two source ontologies are also used as a foundation for BIRNlex.

/wiki/images/1/17/Fish1.png /wiki/images/e/ea/Fish2.png /wiki/images/f/fa/Fish3.png /wiki/images/f/ff/Fish4.png /wiki/images/4/40/Fish5.png /wiki/images/c/c5/Fish6.png
Personal tools