Fig. 3

Representative schema for data-driven classification. Blue nodes (i1–3) are OWL individuals representing clusters of single-cell transcriptomes, while tan nodes (c1, c2) are OWL classes representing cell types. Hierarchical clustering is represented using the transitive subcluster_of relation (objectProperty) to link individuals. Each class is defined by reference to a cluster individual (i), via the relation (objectProperty) as equivalent to (any) cell that has_examplar (value) i. Reasoning via a chain of these two properties (bottom and right sides of the diagram above) is sufficient to infer that c3 has_examplar value i1 and so, combined with the assertion that it is a (type of) cell, fulfils the conditions required to be a subclass of i1.