|
Protein-protein interaction networks (PPIN) are the backbone of cellular function and allow for the identification of several biological processes, such as important pathways and macromolecular complexes. Apart from the evident biological information provided by PPIN, they also provide a series of topological properties such as degree distributions, assortativity, average degree, clustering coefficient, etc. Among many other useful purposes for the use topological information, is the direct comparison of networks either between different species or between networks of the same specie obtained by different experimental techniques. Most studies so far have fallen short of a realistic topological comparison between networks, this mainly due to the fact that only degree distribution have been used for such comparisons. The biggest shortcomings of comparing degree distribution relate to the fact that they only provide part of the information encoded on a network. Furthermore, networks with the same degree distribution can differentiate on the actual network realization. Here we apply a series of rigorous mathematical methods, which we have recently developed, to the topological exploration of several PPIN available in the public domain. We first apply a systematic analytical framework with which to quantify the macroscopic topological structure of PPIN along with the generation of reliable 'null models' (random graphs whose topological features can be controlled precisely). We calculate entropy measures that relate to network complexity. We define complexity according to the uniqueness of a network when compared to an equivalent null model network. Alongside those calculations we also developed methods that determine differences between networks purely based on topological properties and apply those methods to perform a species to species comparison. Preliminary results indicate that the currently available data does not allow species to species comparison; this is manly due to experimental differences but also due to the low sampling seen in current PPIN. Although the availability of datasets with adequate coverage and of high accuracy is lagging, we have now created the necessary tools for a detailed topological characterization and direct network comparisons. |
|