Meeting with Dr. Bultelle
Modelling of enzymatic reaction
 Simulation step might be too big. We need a very small simulation step, given our simulation time.
 Negative concentrations are impossible. Once our concentration goes negative once, the model will just not work properly anymore. So, we have to implement a minimum concentration of c = 0 into our Matlab program.
 Look into the ode45 solver in Matlab and its different options regarding accuracy.
 Tinkercell uses a different algorithm!!!
 It is possible to multiply all quantities (i.e. concentrations) by a constant to get above a certain threshold (don't forget to rescale equations and especially kvalues). However, this increases the margin of error.
Modelling of ComD receptor
 Important questions: Is the receptor working like an enzyme? Does it only work once?
 We need to be very precise when modelling the receptor. (Look into the biochemistry of the receptor.)
 Look into the binding of the receptor, e.g. can CSP unbind under any circumstances.
 Always check and list the assumptions of the model, e.g. for enzymatic reactions: quasisteady state
 Check if the assumptions are met for the ComEComD system (look at the paper!)
 Look into crosstalk and any other possible reactions
Meeting with Dr. Stan
Modeling enzymatic reaction
Having implemented some of the tips suggested by Matthieu (scaling up the numbers and restricting solver to nonnegative values), we still were getting some negative values for our model. As we considered that very wierd and we were quite confident that the equations we have used for our model are reasonable, we tried to get one more opinion on the issue.
Key points:
 Check graphs for conservation of species (but remove the production and degradation term). If there are still oscillations visible, then they are due to production and degradation term or MatLab is doing something wrong
 Our differential equations are stiff (we need to read up on that). Using ode15s solver is more appropriate in such case as compared to ode45 used before. ode15s is designed to deal with stiff differential equations. Apprarently, introduction of constants in differential equations has made our system stiff. Using ode15s removed oscillations completely.
 Increase AbsTol (absolute Tolerance) which indicates values of smallest numbers that MatLab will deal with. However, it takes up more computer power.
Modelling of ComD receptor
 Look at Sontag notes for MAP kinase model  to have an idea how these things are done.
 Look at the mechanism of: phospohrylation, dephosphorylation, 2 component systems
