A biophysicist ponders the application of hidden metric spaces to genetic networks.

Complex networks can be conceptualized as a collection of points or 'nodes' connected by edges that represent their interactions. The structure and logic of these visualized networks allows mathematical modelling to investigate dynamics such as how information propagates through a system. I am particularly interested in gene regulatory — ensembles of molecules and interactions that control gene expression — because of their connection to human diseases such as cancer.

Marián Boguñá and his colleagues suggest that real, observable networks are underlain by geometric frames that contain all nodes, influence topology and guide information-routing decisions. They call these underlying frames 'hidden metric spaces' (M. Boguñá et al. Nature Phys. 5, 74–80; 2009).

In addition to the distance between nodes in the observable network, one can measure similarity between nodes — which can be determined by, for example, how many neighbours they share — and abstract it as a 'hidden distance'. Hidden distances are then used to define the hidden metric space that would place similar nodes closer together, increasing the probability that they are connected and interacting in the network.

A major challenge to applying this framework is explicitly identifying the structure of the hidden metric space for complex networks, for which data sets are often noisy or incomplete. For genetic networks, hidden distances could be abstracted from available data such as tissue-expression profiles. Comparing hidden metric spaces constructed from different data types with known genetic interactions would identify which data are best suited to the process. It will be interesting to see what this reveals in terms of the structure and dynamics of genetic networks.

Discuss this paper at http://blogs.nature.com/nature/journalclub