Statistische Modellierung und Analyse biologischer Netzwerke

• Applicant: Professor Dr. Johannes Berg
• Subject Area: Bioinformatics and Theoretical Biology
• Term: from 2006 to 2010
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 13575000

• Final Report Year: 2010

Final Report Abstract

The central question of this project was: what can we learn from the growing volume of genomic data on the function and evolution of bio-molecular networks? Since the start of this project in 2006, the volume of data in the form of networks – regulatory networks, co-expression networks, protein interaction networks, metabolic networks – has grown severalfold, and novel experimental techniques to probe bio-molecular interactions have been established. In this project, statistical models and methods for the analysis of bio-molecular networks were developed. A key step in in the analysis of biological data (sequences, morphologies, networks, etc.) is the comparison across species: Which elements have changed, which are particularly conserved? In the case of large networks, it is already hard to determine which parts in the network of one species correspond to which parts in the network of another species (the network alignment problem). We have developed concepts and tools for the alignment of networks. In analogy with the widely-used sequence-alignment, a graph alignment maps nodes in one network (e.g. genes in a regulatory network) onto nodes in another network, thus highlighting parts of the networks which are particularly similar to each other, and those parts which have undergone large-scale changes. The graph alignment turned out to be particularly effective where sequence data gives little information, that is short genes and large evolutionary distances. A second line of enquiry focussed on determining regulatory interactions from gene expression data. Given expression levels from many genes (amount of messenger RNA in a sample), can one determine which transcription factors affect the expression of which target genes? We have developed stochastic models of gene expression which allow to determine regulatory interactions from time-course gene expression data. These models are based on a link with out-of-equilibrium physics: expression levels of transcription factors can be considered acting as a ‘force’ on expression levels of target genes. The constant changes in transcription factor expression levels drive expression levels of target genes away from equilibrium. Tracing this connection between transcription factors and their target genes in experimental data allows to infer regulatory interactions.

Publications

• Bayesian analysis of biological networks: clusters, motifs, cross-species correlations, Statistical and Evolutionary Analysis of Biological Network Data, M. Stumpf and C. Wiuf (Eds.)
  J. Berg and M. Lässig
  
• Cross-species analysis of biological networks by Bayesian alignment, PNAS 103(29), 10967-10972 (2006)
  J. Berg and M. Lässig
  
• Statistical modelling and analysis of biological networks, in Complex Systems, J.-P. Bouchaud, M. Mézard, and J. Dalibard (Eds.), Elsevier (2007)
  J. Berg
  
• Dynamics of gene expression and the regulatory inference problem, Europhys. Lett. 82, 28010 (2008)
  J. Berg
  
• From Protein Interactions to Functional Annotation: Graph Alignment in Herpes, BMC Systems Biology, 2:90 (2008)
  M. Kolář, M. Lässig, and J. Berg
  
• Out-of-equilibrium dynamics of gene expression and the Jarzynski equality, Phys. Rev. Lett. 100, 188101 (2008)
  J. Berg
  
• Adaptive gene regulatory networks, EPL 88, 48004 (2009)
  F. Stauffer and J. Berg