Figure 6
From: Structural measures of similarity and complementarity in complex networks
Structural similarity and complementarity in social (19 cases) and biological (9 cases) networks. Scatterplots show calibrated average nodewise coefficients with dashed lines indicating null model expectations based on UBCM (see “Materials and methods” section for details on the datasets and calibration). Colors of the quadrants of the plots indicate configurations of increased values of similarity and complementarity coefficients—high and low (red), low and high (blue), both high (violet). (A) Social networks. Almost all feature high similarity and some also increased complementarity (these are mostly large online social networks). The only case with low similarity and relatively high complementarity is the network of Dutch literary criticism representing relationships of recognition (mentioning other’s work, positively or negatively, in an essay or an interview) within a set of notable literary authors40. (B) Interpretation of similarity and complementarity in terms of homophily, preferences and transitivity. Some social relations, especially those depending on close bonds such as friendship or trust, are often driven by homophily8,9. This implies transitivity of ties and the abundance of triangles due to triangle closure. However, other relations such as recognition or skill-based collaboration23 are based on preferences decoupled from the properties of the ego. In this case two nodes with similar preferences connect to the same neighbors but not necessarily to each other. This leads to what can be called second-order transitivity, which in turn implies quadrangle closure. (C) Biological networks. Most of them feature both increased similarity and complementarity indicating higher structural diversity than in the case of social networks. This is consistent with multiple results reporting the abundance of both triangle and quadrangle based motifs1,2,3. (D) Examples of the relationship between directed motifs often reported for biological networks and undirected motifs used for defining structural similarity and complementarity.
