Executable Symbolic Models of Neural Processes

Background Neuroscience is experiencing explosive growth in detailed high-quality experimental information on neural processes underlying learning, memory and behavior. There is a need for computational models that can manage this outpouring of information, derive knowledge from information, and to generate novel, testable hypotheses. Many current models of neural processes utilize a framework of differential equations. These models tend to exhibit high sensitivity to system parameters requiring accurate measurements of these parameters. Such data are frequently unavailable, leading to difficult solution stability, robustness, and validity problems. Further, the models do not scale easily since they rapidly become intractable as the number of cells incorporated increases. The situation is analogous to that in the biochemical pathway modeling. There, a complementary approach based on a computing formalism called Pathway Logic, implemented by the high-level rewriting-logic language Maude. This approach is being successfully used to model and simulate biochemical signaling pathways.