We consider the core of MesoRD, i.e. the stochastic without the additions made in the latest version, to be relatively stable. This part of the code is also heavily tested and used. However, this does not mean that the code is error free in any way. Please report any errors you encounter, be they in the source code or the documentation. Suggestions for improvements are welcome as well. You can report bugs using the web interfaces at SourceForge, http://sourceforge.net/projects/mesord. The SourceForge pages allow for posting feature requests as well.
Fixed bug in checkpointing and configure script.
The choice of reaction-partners for bi-molecular reaction using neighbouring subvolumes is now scaled with the diffusion rate of each species. See the section called “ Scale dependent association rate constants ” for more details.
Support for microscopic reaction rate constants, a la, [Fange et al. 2010], which makes it possible to reach a spatial resolution on the size of reaction radii of interacting molecules.
Support for truly planar 2D surfaces.
Adherence to the SBML version 2, level 4 ( http://precedings.nature.com/documents/2715/version/1). MesoRD is NOT kept backward compatible with the earlier SBML versions, which results in a number of changes in the model-files. See the section called “ Unit ” and the section called “ Compartment ” for details.
Rewritten CSG module.
Support for the libsbml-4.1
Command-line support on Windows™ platforms.
Some of the items here may be of interest mainly to developers. This list is most likely not complete!
The visualiser class is unstable with respect to optimisations on Windows™, at least when compiled with icc and vc++-8.0. Whether this is due to bad code, bad compiler or an evil operating system is yet unknown.
The CSG module seems to suffer from numerical instabilities under icc optimisation. It seems the problem is related to near-singular matrices. The implementation can certainly be made to be more robust.
Running deterministic simulations for models with many compartments with many different molecule kinds will result in a lot of unnecessary computations.
One final note. I am not Bjarne Stoustrup. Neither is anybody else in the MesoRD development team [4]. If you think that the code looks stupid and could implemented better, you are most likely right. We encourage any improvements. We appreciate your help!
We would like to support SBML in full, with events and functions and everything.
Support for more kinds of subvolumes, like tetrahedrons. This work has been started but is yet not completed.
We would like to be able to simulate 1D and 2D geometries embedded in 3D and to treat diffusion limited reactions correctly in these geometries. (This is not very easy.)
Can we implement other algorithms that gives complementary information, such as the GFRD (Greens function reaction diffusion) algorithm?
We should make an efficient implementation for some parallel architecture.
[4] I know this for a fact, because there are only two developers at the time of writing. (This is still true for version 0.2 and 0.3, although we a are now three developers.)