Fig. 3

Clustering of skin diseases on the basis of phenotype similarity. The phenotypic profiles of each disease in DermO are related to each other through the phenotype ontologies and the strength of the association reflected in the presence and length of the lines connecting them. The algorithm for this is described in detail in Hoehndorf et al. [27]. This example shows close clustering of three types of skin disease, carried out automatically on the basis of the spectrum of text-mined phenotypes. This network is based on phenotype/disease associations established using NPMI [28] with phenotypic similarity determined using Phenomenet [36]