Building networks with microarray data

Multi-scale genetic dynamic modelling II: application to synthetic biology: an algorithmic Markov chain based approach

We model in detail a simple synthetic genetic clock that was engineered in Atkinson et al. (Cell 113(5):597-607, 2003) using Escherichia coli as a host organism. Based on this engineered clock its theoretical description uses the modelling framework presented in Kirkilionis et al. (Theory Biosci. doi: 10.1007/s12064-011-0125-0 , 2011, this volume). The main goal of this accompanying article was to illustrate that parts of the modelling process can be algorithmically automatised once the model framework we called 'average dynamics' is accepted (Sbano and Kirkilionis, WMI Preprint 7/2007, 2008c; Kirkilionis and Sbano, Adv Complex Syst 13(3):293-326, 2010). The advantage of the 'average dynamics' framework is that system components (especially in genetics) can be easier represented in the model. In particular, if once discovered and characterised, specific molecular players together with their function can be incorporated. This means that, for example, the 'gene' concept becomes more clear, for example, in the way the genetic component would react under different regulatory conditions. Using the framework it has become a realistic aim to link mathematical modelling to novel tools of bioinformatics in the future, at least if the number of regulatory units can be estimated. This should hold in any case in synthetic environments due to the fact that the different synthetic genetic components are simply known (Elowitz and Leibler, Nature 403(6767):335-338, 2000; Gardner et al., Nature 403(6767):339-342, 2000; Hasty et al., Nature 420(6912):224-230, 2002). The paper illustrates therefore as a necessary first step how a detailed modelling of molecular interactions with known molecular components leads to a dynamic mathematical model that can be compared to experimental results on various levels or scales. The different genetic modules or components are represented in different detail by model variants. We explain how the framework can be used for investigating other more complex genetic systems in terms of regulation and feedback.

Bioinformatic analysis of molecular network of glucosinolate biosynthesis

Glucosinolates constitute a major group of secondary metabolites in Arabidopsis, which play an important role in plant interaction with pathogens and insects. Advances in glucosinolate research have defined the biosynthetic pathways. However, cross-talk and interaction between glucosinolate pathway and other molecular pathways are largely unknown. Here three bioinformatics tools were used to explore novel components and pathway connections in glucosinolate network. Although none of the software tools were prefect to predict glucosinolate genes, combination of results generated by all the tools led to successful prediction of all known glucosinolate genes. This approach was used to predict new genes in glucosinolate network. A total of 330 genes were found with high potential to relate to glucosinolate biosynthesis. Among them 64 genes were selected to construct glucosinolate network because their individual connection to at least one known glucosinolate gene was predicted by all the software tools. Microarray data of candidate gene mutants were used for validation of the results. The mutants of nine genes predicted by glucosinolate seed genes all exhibited changes in the expression of glucosinolate genes. Four of the genes have been well-known to functionally interact with glucosinolate biosynthesis. These results indicate that the approach we took provides a powerful way to reveal new players in glucosinolate networks. Creation of an in silico network of glucosinolate biosynthesis will allow the generation of many testable hypotheses and ultimately enable predictive biology.