In silico evolution of gene regulatory networks for the optimization of metabolic networks

From Master Projects
Jump to: navigation, search

has title::Design of self-optimizing biochemical networks
status: finished
Master: project within::Bioinformatics
Student name: student name::Timo Maarleveld
Start start date:=2010/12/01
End end date:=2011/05/31
Supervisor: Herbert Sauro
Second reader: has second reader::Sanne Abeln
Company: has company::University of Washington
Poster: has poster::Media:Media:Posternaam.pdf

Signature supervisor




Gene networks mediate the expression of genes with transcription factors as inputs. At the level of single genes, transcription factors modulate gene activity either in an individual or cooperative fashion. The topologies of gene networks - the identity of transcription factors and the location and sequence of the cis-regulatory response sequences - are being discovered at a rapid pace from experimental data. To come to understand the kinetics, functions, and dynamics of gene networks, we have to take a systems perspective. Here, we must appreciate the nature of transcription factor regulation, transcription factor inputs, and the biological role of the protein products of the regulated genes by integrating gene networks with their input and target networks, which could play either a signaling or metabolic role. This represents one of the main challenges in gene network research at the moment.


Small molecules, e.g. metabolic intermediates, or the activity of kinases and phosphatases function as the regulators of the transcription factor input layer of gene networks. The outputs of gene networks are the levels of protein products. The relationship between the levels of the regulators and the protein products together form the input-output characteristic of a gene network.

This characteristic can either be measured or computed from a kinetic model of the associated the metabolic or signaling network that acts both as the input and the target of the gene network. Here we take an optimization approach, we assume that the target network is optimized for some function(-s) and that the enzyme levels available for this network obeys some constraint.

Using a nonlinear optimization algorithm one can now predict the enzyme levels that optimize the network function given the constraint. Here, existing optimization algorithms could be used, but our objective is to develop clever algorithms that optimize the network given the constraint.Those enzyme levels are the output of the gene network that has received information from the target network’s internal state, e.g. the levels of active signaling proteins or metabolite concentrations.

Thus, the optimization procedure predicts the optimal input-output characteristic of the gene network. Gene network structures with the desired input-output characteristics can be evolved in silico from a limited number of gene network building blocks and allowable parameterizations.