MesoRD will build the geometry of the system using fixed size sub volumes. Theoretically, there are several differently shaped sub volumes one could use for this construction. Currently, MesoRD only knows about one kind of sub volume: the cube.

To give physically reasonable results the size of the cubic subvolume must meet some criteria that are described in Chapter 6, Rates and state changes in MesoRD . However practically the most important criteria is the result of the simulations should not change if the size of the artificial subvolume changes a little. Thus we suggest that you run your simulations with doubled and halved subvolume size to make sure the result does not depend on the artificial discretisation.

Any simulation is typically run until a certain termination criterion is met. One could imagine implementing all kinds of complicated convergence criteria. We have been very unimaginative with our termination criteria: a simulation is run either a certain number of iterations or a certain number of simulation seconds. Simulation is terminated when either of these conditions are met. There is one trick with setting the maximal number of simulation iterations: it may very well occur that the number of simulations one would want to run will not fit into a C long integer type. This may turn into a problem, especially on 32-bit architectures. Thus, the maximal number of iterations must be specified using a multiplier and a scale. The maximal number of iterations is then interpreted as the multiplier times ten to the power of the scale.

A checkpoint is a dump of the current state of the system. There are two types of checkpointing in MesoRD: Full state checkpointing and compartment state checkpointing. In full state checkpointing the whole state of the system, i.e. the number of molecules in the different subvolumes is append to several files, as further discussed in the section called “ Making Sense of MesoRD Output ”. If the whole state is saved, it is possible to restart the a simulation from this point in time. In compartment state checkpointing the number of molecules of different species are saved for each compartment.

Note that if both its and secs are greater than zero, MesoRD will checkpoint whenever its iterations or secs seconds have passed since the last checkpoint. Since checkpoints can consume much disk space and are quite expensive in terms of running time, this may not be desirable.

In stochastic simulations there will be a high number of empty subvolumes, especially if the number of molecules is low, therefore MesoRD has the possibility to write the full state checkpoints in a sparse format. The format of the files is described in the section called “ Making Sense of MesoRD Output ”. Note: The deterministic simulation mode has no support for the sparse format, weird this might happen if you use them together.

MesoRD contains a few visual elements. The most prominent of these is the visualiser, which enables real-time viewing of the system. Using the visualiser it is possible to see whether the simulation does what one would expect and it thus enables debugging of model definitions. The visualiser is not intended for any accurate analysis of simulation results. The visualiser is also not necessary for the simulation per se, and since it consumes a fair share of available resources, its use is optional. If you have a poor graphics card, you are therefore recommended to use the visualiser only for debugging the SBML file. To make optimal use of the visualiser you should have a fast dual processor machine with a high end graphics card.

The least prominent visual element is the progress bar. The progress bar displays the current iteration number and the the current simulation time. The progress bar also enables clean, premature termination of a simulation.

MesoRD will run a mean-field simulation of the model when the deterministic mode is selected. There are a few numerical integration methods to choose among. All of them will require a time step for each iteration. The implicit methods will in addition require some parameters for the equation solver.

A simulation can be resumed from an old full state checkpoint assuming that the SBML .xml file is read from a directory where a previous simulation has been run with the same geometry and discretisation and that all old checkpointing files are still present. (default this feature is turn off).

The output files in the directory will overwrite the old simulation data that is used for to define to start state, i.e. make a copy of your old simulation data before starting from an old checkpoint.

Note: You cannot start a deterministic simulation from a previous stochastic simulation that has been stored in the sparse format. You cannot start a stochastic simulation from a previous deterministic simulation, since there is not an integer number of molecules in the different subvolumes. If you try this there will be no warnings, just strange output from your new simulation.