Email updates

Keep up to date with the latest news and content from BMC Bioinformatics and BioMed Central.

Open Access Research article

Bayesian statistical modelling of human protein interaction network incorporating protein disorder information

Svetlana Bulashevska1*, Alla Bulashevska1 and Roland Eils12

Author Affiliations

1 Theoretical Bioinformatics Department, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, 69120 Heidelberg, Germany

2 Department of Bioinformatics and Functional Genomics, Institute of Pharmacy and Molecular Biotechnology (IPMB) and Bioquant, University of Heidelberg, Germany

For all author emails, please log on.

BMC Bioinformatics 2010, 11:46  doi:10.1186/1471-2105-11-46

Published: 25 January 2010

Abstract

Background

We present a statistical method of analysis of biological networks based on the exponential random graph model, namely p2-model, as opposed to previous descriptive approaches. The model is capable to capture generic and structural properties of a network as emergent from local interdependencies and uses a limited number of parameters. Here, we consider one global parameter capturing the density of edges in the network, and local parameters representing each node's contribution to the formation of edges in the network. The modelling suggests a novel definition of important nodes in the network, namely social, as revealed based on the local sociality parameters of the model. Moreover, the sociality parameters help to reveal organizational principles of the network. An inherent advantage of our approach is the possibility of hypotheses testing: a priori knowledge about biological properties of the nodes can be incorporated into the statistical model to investigate its influence on the structure of the network.

Results

We applied the statistical modelling to the human protein interaction network obtained with Y2H experiments. Bayesian approach for the estimation of the parameters was employed. We deduced social proteins, essential for the formation of the network, while incorporating into the model information on protein disorder. Intrinsically disordered are proteins which lack a well-defined three-dimensional structure under physiological conditions. We predicted the fold group (ordered or disordered) of proteins in the network from their primary sequences. The network analysis indicated that protein disorder has a positive effect on the connectivity of proteins in the network, but do not fully explains the interactivity.

Conclusions

The approach opens a perspective to study effects of biological properties of individual entities on the structure of biological networks.