Seminar 20 April
From the Manchester Bio-Modelling Network
The BioModelling Network organised a Seminar on Friday the 20th of April from 12:00 to 13:00, in MIB (Lecture Theatre).
Speaker 1: Dr Steve Wilkinson
Title: Good ways to build bad models
Abstract
Building ODE models of signalling and metabolic networks can be frustrating because, although the stoichiometric structure of the network might be well characterised, the values for many of the parameters (rate constants, initial concentrations) are not known. In addition, measured time series data with which to estimate the unknown model parameters are often rather scarce leading to a very under-determined system. In our modelling efforts we have chosen to make the best of what little data is available by preserving the structural detail of the network and, starting from best guesses for each parameter, tune them until the model fits the measured data while keeping as close as possible to our initial best guesses. At least we then have a quantitative model that fits the data – even though this model is almost certainly wrong in terms of its individual parameter values. We argue that it is better to have a ‘bad’ model than no model at all since it kick starts the iterative loop of experimental testing and model refinement that leads to better models and greater understanding.
Slides of the seminar here.
Speaker 2: Dr Nils Bluethgen
Title: Reverse engineering of a negative feedback in MAPK signalling
Abstract
Transcriptional regulation is the major means to orchestrate the temporal organization of the protein network in the living cell. Decay rates of proteins and mRNAs determine the timescale on which this network can be modulated. Considering that e.g. liver cells spend more than a third of their energy budget on protein turnover, considerable evolutionary pressure to optimize mRNA and protein stability can be expected, which has to be reflected in the topology of the transcriptional network. Recent genome-wide measurements of mRNA half-lives open the door to investigate this hypothesis in great detail.
We report principles behind the design of transcriptional feedbacks in the best studied signal transduction pathway in the mammalian cell, the mitogen activated protein kinase (MAPK) cascade. By integrating time-resolved expression data, sequence data, and functional annotations, we reconstruct the transcriptional feedback loops that shape the activity of the cascades. We find that activation of the MAPKs induces a phalanx of phosphatases that feed back on the activity of the kinases and identify the transcription factors involved. By means of mathematical modeling, we show that induction of phosphatases rather than the down-regulation of the kinases provides a fast yet energy efficient design of deactivating and reprogramming the MAPK signal transduction network. The prediction of this model is that DUSPs decay rapidly in contrast to the kinases which is confirmed by experimental data.
As an example of this feedback regulation we further investigate a specific feedback, the induction of DUSP6 by the classical MAPK pathway consisting of Raf/Mek/Erk. We utilize an inducible oncogenic RAS to stimulate the cascade and monitor phosphorylation of Erk as well as the expression of a few hundred target genes. We construct matematical models and perform model selection on the data, and find that DUSP6 indeed constitutes a negative feedback loop that rapidly down-regulates the activity of the MAPK Erk and mediates its adaptation. Furthermore, the model predicts that Erk-activation is ultrasensitive, which we confirm by measuring Erk-activity quantitatively in single cells. We show that incoherent feed-forward loops in the induction of DUSP1 and DUSP6, and ultrasensitization additionally contribute to the speed and modularity of the response.
Slides of the seminar here.
