Network motif identification and structure detection with exponential random graph models

• Main Authors: Munni Begum, Jay Bagga, Ann Blakey, et al.
• Format: Article
• Language: English
• Published: International Academy of Ecology and Environmental Sciences 2014-12-01
• Series: Network Biology
• Subjects: biological networks; network motifs; transcriptional regulatory network; graphical models; exponential random graph models; Markov Chain Monte Carlo algorithms
• Online Access: http://www.iaees.org/publications/journals/nb/articles/2014-4(4)/network-motif-identification-and-structure-detection.pdf

Local regulatory motifs are identified in the transcription regulatory network of the most studied model organism Escherichia coli (E. coli) through graphical models. Network motifs are small structures in a network that appear more frequently than expected by chance alone. We apply social network methodologies such as p* models, also known as Exponential Random Graph Models (ERGMs), to identify statistically significant network motifs. In particular, we generate directed graphical models that can be applied to study interaction networks in a broad range of databases. The Markov Chain Monte Carlo (MCMC) computational algorithms are implemented to obtain the estimates of model parameters to the corresponding network statistics. A variety of ERGMs are fitted to identify statistically significant network motifs in transcription regulatory networks of E. coli. A total of nine ERGMs are fitted to study the transcription factor - transcription factor interactions and eleven ERGMs are fitted for the transcription factor-operon interactions. For both of these interaction networks, arc (a directed edge in a directed network) and k-istar (or incoming star structures), for values of k between 2 and 10, are found to be statistically significant local structures or network motifs. The goodness of fit statistics are provided to determine the quality of these models.