The NoTremor Vision
The computational modelling and simulation work-plan of NoTremor is based on two principles. First that it should partition off model description from model simulation, wherever possible so as to be able to perform processing in a modular manner, and second, that it should aim to re-use existing simulators, codes and frameworks wherever possible, whether open-source or available in the consortium.
The first principle is related to the notion of domain specific model description languages which have become widely used in many research areas to facilitate design and specification of large-scale simulation models. Thus, the models can be described in a model-description-language (MDL) which doesn't actually simulate anything, but is a precise description (usually in some .xml extension) of components, signal flows, and dynamics. This MDL file is then compiled down to a simulation platform in a user transparent way. Brain modelling is no exception, and new tools have been recently developed which promise a step change in our ability to build large models that can take full advantage of next generation hardware. NineML is one such example and is the outcome of work by the International Neuroinformatics Coordinating Facility (INCF). Recently, USFD has enlarged the NineML framework to include rate coded models, provided a variety of simulator support, and front end GUI tools for model construction.
More specifically in relation to NoTremor, the tools have been validated with a use-case comprising the implementation of a biologically accurate model of the striatal microcircuit. The new tool set (SNiNeML ) supports compilation to C++, Python and Matlab, and also a ‘sideways’ translation to PyNN (an early and somewhat limited neural MDL) thereby supporting simulation engines like Brian, NEURON and NEST. A similar approach will be also adopted for the biomechanical simulation. The model description will be also provided by MDL, while the potential use of the VERITAS virtual user models, described using OWL and UsiXML format that encoded motor impairments could be used for NoTremor.
Practical realization of the second principle (code re-use) is predicated on being able to integrate running processes defined by a variety of existing codes (and even complete simulation engines). These may all be running at different time steps and represent a heterogeneous set of objects (e.g. neurons, muscles, data visualization streams). This integration is the remit of so-called middleware and, recently, USFD have developed such software for biological simulations dubbed BRAHMS. This may be considered as software ‘glue’ which allows processes running at different rates to be coordinated and communicate with each other in a well orchestrated way. USFD will use this at a comparatively fine level of granularity to orchestrate their neural simulations. CERTH and UPAT will use the inbuilt process managers of OpenSIM and the VERITAS task and multi-body dynamics framework. The optimal integration and communication link among these two processes will be examined in the respective workpackage. These principles and their use in NoTremor are illustrated in the Figure below. The brain-basal ganglia models and simulator on the left and the neuromuscular-biomechanics models and simulator on the right will communicate through BRAHMS as previously described, while the simulation results will be presented to the expert via intuitive visualizations through the NoTremor analytics module.