Ontology-assisted keyword search for NeuroML models

Abstract

NeuroML is an extensible markup language for describing complex mathematical models of neurons and neuronal networks. NeuroML is unique in its modular, multi-scale structure -- not only can entire NeuroML models be exchanged, but subcomponents of these models that correspond to neuroscience objects, like channels or synapses, also can be shared and reimplemented in a different model. This paper presents the design, implementation, and evaluation of an ontology-assisted search for NeuroML models. Specifically, the paper describes the design of the system, including the database that stores the modular NeuroML models and the architecture of the Web-based search (neuroml-db.org). The implementation takes advantage of the nested structure of NeuroML models and the NeuroLex ontology for neuroscience to provide additional semantic information to enhance the search. In addition to NeuroLex terms that may exist in model metadata, this initial implementation takes advantage of several semantic relationships provided by the NeuroLex ontology: Is_part_of, Located_in, and Neurotransmitter. An evaluation of the system illustrates its effectiveness both for functionality and performance, covering various types of searches broken down by keyword searches over the database and ontology searches using the semantic relationships.