On network entropy and bio-interactome applications

Networks represent a main methodological instrument in systems biology applications. In particular, modularity is widely investigated in the attempt to elucidate the regulative and correlative nature of gene and protein associations, respectively. However, modular maps are only approximate representations due to two main factors. First, the resolution spectrum that has to be covered is wide and method-dependent. Second, the randomness underlying network dynamics and influencing them through fluctuations and system perturbations, is difficult to measure. We investigate both aspects by an application to the yeast protein interactome network, and suggest that a non-extensive characterization of entropy may play a role for elucidating both random and biological variation.

Research highlights
► Protein interactomes are complex systems that require further understanding and methodological treatment, especially from a probabilistic approach. Uncertainty is part of such networks and needs to be accounted for by developing ad hoc inference approaches.
► In systems biology applications, modules are not completely defined objects (as discussed in the paper).
► Modularity is a well-known domain of research that presents challenges from both numerical and algorithmic aspects (the resolution limit problem) and also from the interpretation of results.
► Non-extensive entropy plays an important role for network studies, and this work aims to contribute with both theoretical rationale and numerical evidence obtained through application in protein interactomes.