Large numbers of cell and network models in the current literature have been developed by implementing them in a variety of specialist simulators. These models illustrate phenomena of general interest, but this approach is often ignored by the wider experimental neuroscience community. There are many reasons for this, one being that to use the models effectively, the scripting language of the simulation environment has to be learned, which can be time consuming. Also, if two or more cell models for a particular brain area have been implemented on different platforms, it can be a major task even for an experienced modeler to convert these to a single simulator, which is necessary before a composite model can be built. We have recreated a number of published cell models from various brain systems in neuroConstruct and intend to add more cell types to this library in the future. The simulator independent representation of models in neuroConstruct and their control and potential modification though a GUI, without having to write any specialist simulator code, facilitates the accessibility of these models and therefore permits non-programmers to evaluate the model structure and predictions in detail. This increased accessibility for a wider range of neuroscientists is a key feature of neuroConstruct as it allows for continuous feedback and testing of the model predictions in experiments, which is necessary for the ongoing development and improvement of computational models. Moreover, cell models can be designed de novo from new experiments and integrated into increasingly larger composite models. It is hoped that the facilitated interchange of models between experimentalists and theoreticians should lead to the development of models that reflect a broad consensus view of the cellular basis of brain function.
The greater availability of models of healthy neuronal systems will also benefit research into the underlying causes of neurological diseases. As models become more biologically realistic, have been open to scrutiny by a wide range of neuroscientists and have been shown in a wider range of scenarios to reproduce aspects of cellular behavior, their usefulness for exploratory research increases. This is the hallmark of a good model: that it can make experimentally testable predictions, which are not directly linked to the underlying assumptions used when building the model. For example, the effects on a normally behaving cell of deletion of a channel can be quickly checked by copying and altering the cell model in neuroConstruct and comparing the new cell behavior to experimental data from the cell in the presence of the appropriate blocker. This cell can be switched into the network model and the effect on the overall behavior examined. Also, changes in the network topology, e.g. deletion of a certain cell type or increase/decrease in connectivity, can be implemented, the network generated, visually verified and the simulation run, all in a matter of minutes. This quick turn around time between implementing a pathological change in the cells or network and seeing the effect is aimed at reducing the 'time to insight' when using neuronal models for research on neurological disorders. For example, lesions to parts of the modeled network can support research into the causes of symptomatic epilepsies, and changes to the properties of channel models that reflect those occurring following mutations could help shed light on the mechanisms of channelopathy-induced idiopathic epilepsies. Moreover, different forms of pharmacological and non-pharmacological treatment could initially be tested in a model implemented in neuroConstruct, including GABAa receptor modulators such as benzodiazepines, or even the potentially beneficial effect of removal of a cell class or a part of the network.
The greater accessibility to computational models that is provided by neuroConstruct not only facilitates research into the physiology of brain structures, it also provides an advanced tool for teaching. Although neuroConstruct in its present form requires some background in neurophysiology and some basic computing skills, the GUI based control of the model parameters and the advanced visualization of the effect of parameter changes are useful for students who want to understand principles underlying the complex interactions in single neurons and networks. As part of a neuroscience course, structured sets of projects in neuroConstruct can provide examples of the current understanding of the physiology of different brain structures and highlight how their physiology is affected by different pathologies.
Was this article helpful?