My attempts to be useful in computational neuroscience have fallen far short. Perhaps I should have taken to heart some wisdom on the philosophy of modeling before I started dumping a lot of guesswork into a GPU simulation. Here is how I would summarize where I went wrong :
Models with many unknown free parameters should be considered no better than well phrased thought experiments. Without quantitative measurements of all model parameters and verification of the correctness of all model components, the emergent behavior of a complex model can only be a best guess at the qualitative behavior of a system. Such behavior can often be inferred by simple contemplation of the system in question. Therefore, complex modeling should be used when
1 : The model parameters are known with accuracy, and the simulation can be used in leu of in-vivo experiments.
2 : Investigating plausible mechanisms for a specific set of qualitative observations, understanding that multiple mechanisms can generate the same qualitative behavior, with the hope of suggesting experiments to clarify the set of mechanisms involved.
3 : If the model has sufficient mathematical elegance and abstractness to warrant investigation as a theoretical object detached from the biological system under investigation.
A complex model should not be created for the singular purpose of dumping everything you can find in the literature into the computer and watching it go. Specific goals are essential.