Bridging the gap between Computational Neuroscience and Systems Biology modelling

p { margin-bottom: 0.21cm; }a:link { }

The increasing use of computational models as a tool for understanding complex neuronal phenomena has led to a number of initiatives that promote sharing and reuse of detailed neuroscience models. For example, ModelDB (http://senselab.med.yale.edu/modeldb/) is a resource dedicated to the exchange of computational neuroscience models which have appeared in publications, and NeuroMorpho (http://neuromorpho.org) is an online repository of digitally reconstructed neuron morphologies which can be used for detailed single cell simulations. Additional efforts focus on making models more accessible by promoting use in multiple simulators and operating systems. Examples are NeuroML (http://www.NeuroML.org), which currently provides a model description language for representing 3D networks of detailed conductance based neuronal models, and PyNN (http://neuralensemble.org/PyNN), which is a Python API for creating large scale networks of simplified neuron models on multiple simulators.

Parallel activities in the field of Systems Biology have focused on creating standard languages for describing models of biochemical networks in order to promote model exchange and accessibility. SBML (http://www.sbml.org) is widely supported with over 200 compliant tools and an actively developed multi-language library (libSBML), and there is a growing family of associated languages which interact well with SBML (e.g. SED-ML for simulation specification, SBGN for creating network interaction diagrams). The BioModels database (http://www.ebi.ac.uk/biomodels-main) is a resource containing hundreds of published models that have been converted to SBML. CellML (http://www.cellml.org) is another language in this field with associated tools, a multi-language API and a repository of curated models (http://models.cellml.org). Both SBML and CellML can be used for more general applications beyond biochemical signalling pathways, and there are examples of spiking cell models available in both formats.

Both of these communities would benefit from access to the tools and models of the other field, and in the latest v2.0, NeuroML has been extended to allow descriptions of models that span both Computational Neuroscience and Systems Biology. This updated version of NeuroML is designed to interact with a new underlying framework that can be used to describe the behaviour of the core components of neuronal models (e.g. cells, ion channels, synapses). This language, the Low Entropy Model Specification (LEMS), can be used to describe the dynamical behaviour of component types, and a reference implementation is available (http://www.NeuroML.org/lems) to simulate these or map them into widely used simulator formats (e.g. NEURON).

Abstract cell models such as the Izhikevich neuron model or adaptive exponential integrate and fire model can be specified in a very compact format in NeuroML 2.0 and mapped via the LEMS framework to SBML for use on one of the many compliant simulators (e.g. CellDesigner, Copasi). Alternatively, SBML models from the BioModels database (e.g. whole cell models, synapse models) can be imported into LEMS and used beside NeuroML components to create complex multiscale models, which can be mapped to simulators like NEURON. We will demonstrate examples of these interactions, along with preliminary work that incorporates CellML models into this framework. The modelling tool neuroConstruct (http://neuroconstruct.org) can be used to link these different model components and provides a user-friendly graphical interface for creating complex 3D models using mixed model specification formats. This approach will increase interaction between these relatively separate fields, widening the range of models available to researchers and improving simulator technologies in both.