Biological knowledge evolves in a quick way whereas more people with other backgrounds (e.g., bioinformaticians, computer scientists, etc.) that help in biological research require detailed knowledge on biomolecular processes in order to understand the data they need to analyse. To solve this problem, in this paper we propose a multimodal knowledge representation using graphical diagrams representing biological knowledge, a natural language elucidation of the content of the graphical diagrams, linking the graphical elements to ontology instances; and a graph for visualising the ontology.
Biological research produces highly diverse information types [
GREEKC [
In order to satisfy these requirements, we propose a model that is ontology-based
and follows a view on ontologies that emphasises them as artefacts for knowledge
sharing within and across domains. Ontologies are therefore seen as formal descriptions of
the characteristics of biological entities (e.g., molecules, organisms, cell components,
processes, qualities, etc.) [
Bio-ontologies are normally restricted to T-Boxes, i.e., axiomatic descriptions of properties that universally hold for all particulars that instantiate a certain type (e.g., that all chromosomes are constituted by DNA). However, T-Boxes are neither sufficient in granularity and expressiveness nor appropriate to fulfil our educational goals. Traditionally, such information has been conveyed by texts and by graphical diagrams, albeit in a rather informal way. GREEKC proposes completing the picture by adding formalontological descriptions as a means to create and disseminate knowledge commons.
In this paper, we use the domain of gene regulation to describe an example of KR model that fulfils the above-mentioned requirements.
In order to build our model, we are assuming that a well-defined T-box exists, with universally agreed meanings (e.g., axioms) and sources such as domain ontologies connected to foundational ontologies as their building blocks. Once we have these blocks, the representation of prototypical examples, such as “transcription factor activity” from Gene Ontology is then expressed as A-box entities (prototypical instances) using: (i) elements of graphical diagrams, having appropriate labels and ideally having interactive functionality that links its graphical elements to the instances they represent in (ii) an ontology that provides universal descriptions in an OWL T-Box, instantiated by A-box entities and expressions that formally describe the processes depicted in the graphical diagram; (iii) a natural language elucidation of (i) (Figure 1); and (iv) a graph visualisation of (ii) (Figure 2).
For (i) we proposed a pre-existing prepared set of diagrams, i.e., graphical depictions
of biological processes (Figure 1), supplied by the Norwegian GREEKC partner Astrid
Laegreid. In order to implement (ii) and (iv) we were using Noctua [
Our Noctua models rely on the Gene Ontology (GO) as domain ontology. They are
centred on a particular molecular activity class from the GO Molecular Function (MF)
ontology, represented as a prototypical OWL instance. MF annotations are connected to
provide the context in which that particular molecular function occurs. All connections
within a GO-CAM model are relations as OWL object properties from the OBO
Relations Ontology. GO-CAM models can be created using the graphic interface of the
Noctua website. The first step to create the triplets was to analyse each statement in the
textual description in order to extract the suitable GO terms by searched identifiers and
keywords. Next, the relations between MF and other GO elements (precisely, instances of
GO classes, particularly from the cellular component (CC) and biological process (BP)
ontologies were added. Finally, instances from Sequence Ontology (SO) [