Domain-Specific Modeling as a Pragmatic Approach to Neuronal Model Descriptions

摘要

Biologically realistic modeling has been greatly facilitated by the development of neuro-simulators, and the development of simulator-independent formats for model exchange is the subject of multiple initiatives. Neuronal systems need to be described at multiple levels of granularity, and compared to other such multi-level systems they also exhibit emergent properties, which are best described with computational and psychological terminology. The links between these levels are often neither clear in terms of concepts nor of the underlying mathematics. Given that modeling and simulation depends on explicit formal descriptions, we argue that rapid prototyping of model descriptions and their mutual relations will be a key to making progress here. Here we propose to adapt the paradigm of domain-specific modeling from software engineering. Using the popular Eclipse platform, we develop the modular and extensible NeuroBench1 model and showcase a toolchain for code generation, which can also support, mediate between, and complement ongoing initiatives. This may kick-start the development of a multiplicity of model descriptions, which eventually may lead to ontologically sound multi-level descriptions of neuronal systems capturing neuronal, computational, and even psychological and social phenomena.