Integration into the Basic Formal Ontology

The [Basic Formal Ontology] (http://www.ifomis.org/bfo) (BFO) is an upper-level ontology that describes material entities bounded in space and time. BFO is currently used in several biomedical ontologies including the NIFSTD ontologies.

Although a model can be considered as a material entity through it implementation or the publication that describes it, models cannot really be bounded in space. However, models aim to describe material entities (real neurons, synapses, or part of the brain). IAO has worked on a similar problem to deal with the representation of information content entities such as data label or document. A particular relation, the “is about” relation has been created in IAO to link a material entity named an information bearer and generically_dependent_continuant named information content entity.

For the integration of CNO into BFO we used similar approach, considering that a model “is about” a material entity. Figure 4 shows examples of such links.

Figure 4: Integration into BFO: how to relate a mathematical model to the corresponding biological entity?

Using this particular assumption we then integrated CNO upper classes into BFO classes. Figure 5 shows the result of this integration.

Figure 5: Integration of CNO upper classes into BFO

In BFO, a material_entity is an independent_continuant. As we assume that a model depends on a material_entity, CNO classes are considered as dependent_continuant. By definition, a dependent_continuant is “a continuant [snap:Continuant] that is either dependent on one or other independent continuant [snap:IndependentContinuant] bearers or inheres in or is borne by other entities”.

As expressed in the definition, a dependent_continuant can be of two sorts: either a generically_dependent_continuant that depends on an independent_continuant bearer or a specifically_dependent_continuant, which inheres in or is born by other entities.

CNO classes describing the model and its structure are generically_dependent_continuants as they describe and therefore depend on material entities such as a neuron, a brain region, ionic channel, neuron subcellular entities, neuron morphology, … In contrast, model qualities and role are specifically dependent continuants.

This sounds quite complicated so here is an example. Let’s consider a neural network model with a recurrent connectivity pattern defined by a particular connectivity rule. The recurrent connectivity pattern is a quality of the model (as the red color would be for a tomato), which is born by a particular connectivity rule. Therefore connectivity rule is a generically_dependent_continuant (which depends on the biological network connectivity) and the recurrent connectivity pattern is a specifically_dependent_continuant and more precisely a quality.

Specifically_dependent_continuants can be either a quality as we just saw or a realizable_entity. Realizable_entities can be a role, a disposition or a function. In the design of CNO, we faced the issue of representing parameters and variables in the ontology. Parameters and variable represent physical qualities such as voltage for instance. In the design of a model, nothing prevents to have voltage as a parameter or a variable. This means that if we create two classes to represent model parameters and model variable, voltage should be a subclass of both classes. To prevent this redundancy, we created the class elementary model component that will contain the references to physical qualities and allow for a particular elementary model component to be either a variable or a parameter depending on the model (e.g. the membrane voltage could be a variable in the sub-threshold equation of an IAF or could be a parameter in some other model). For this, we considered that being a parameter or a variable is a role played by an elementary model component in the model structure (see figure 6).

Figure 6: Using the structure of BFO to define parameters and variables

We thus created a subclass of role named elementary model component role, which contains the subclasses parameter and variable.

###Next - Importing multiple terms from multiple external ontologies