Computational tools for simulation of stochastic ion channel activity in neurons

Lead Research Organisation: University of Edinburgh
Department Name: Royal (Dick) School of Veterinary Scienc


A fundamental goal of modern biology is to understand how the physical and behavioural characteristics of living organisms arise from components, such as cells and molecules, which are often too small to be seen with the naked eye. Considerable progress has been made towards determining how the physical properties of living organisms are specified by their genetic code, which is contained in individual molecules of DNA. By contrast, we understand much less about the physical principles that govern human or animal behaviour. For example, although it is clear that communication between nerve cells is a key component of brain function, the appropriate level of physical detail at which nerve cells must be understood to fully account for human or animal behaviour is far from clear. Most nerve cells have ornate branching structures, called axons and dendrites, which play fundamental roles in processing of information in the brain. In a single nerve cell these structures may contain well over a million ion channels, small molecules that determine how the cell processes information. While in the past neuroscientists have generally only considered how the average activity of this large umber of ion channels influences the function of nerve cells, recent evidence suggests that fluctuations in the activity of individual ion channels may be a critical determinant of nervous system function. Yet, we have few clear insights into how this basic property of ion channel function affects information processing in the brain. One promising approach to this problem is to develop computer models to simulate ion channel activity. However, at present accurately simulating the activity of each ion channel in complex neuronal structures is a formidable task, and it has therefore been difficult to explore how fluctuations in the activity of individual ion channels influences brain function. The goal of the proposed study is to develop new tools to efficiently simulate models of neurons or neuronal circuits that explicitly simulate the activity and location of individual ion channels. These tools will take advantage of recently developed computational algorithms, together with advances in computer science and methods for parallel computing, to reduce the time required for simulation of these models by greater than 100 fold. To facilitate compatibility with other widely used software, the tools will build on current community standards for specification of neuronal models and will be made freely available for download by other researchers or interested parties. Development of these new computational tools will enable new and fundamental questions to be addressed. For example, what particular aspects of neural information processing are most sensitive to fluctuations in the activity of individual ion channels? Do these fluctuations impair neural function, for example by introducing noise, or do they increase the computational power of neural circuits, for example though stochastic resonance effects? If they impair neural function then what mechanisms have evolved to counteract this effect? Conversely, if they confer benefits, then how are these advantages optimized in biological systems? The proposed project will prime new areas of research in the principal investigators laboratory that will aim to address these questions. More generally it will provide a new set of tools, of general use to the wider research community, that may lead to a better understanding of the relationship between the properties of single ion channel molecules, computations carried out by neural circuits and the behaviour of living organisms.

Technical Summary

Recent work suggests that stochastic gating of individual ion channels profoundly affects the computational properties of neurons, but with presently available modeling tools it is difficult to explore the implications of these insights for computations carried out by neurons with complex axonal or dendritic architectures. To overcome these obstacles I propose to develop general-purpose software tools that will substantially reduce the time required for development and simulation of morphologically complex neurons containing stochastic ion channels throughout their axonal and dendritic membranes. To achieve efficiency gains of > 20 fold compared with the current best available solutions we will develop a computational core that uses new versions of the tau leap algorithm for simulation of ion channels. Further efficiency gains will be obtained by enabling the core to run simulations in parallel on multiple processors. To maximize compatibility with existing and future software, the tools to be developed will build on standard formalisms for representation of neuronal models and will also include a version of the software core compatible with the Genesis simulation environment. The time critical components of the core will be developed in Fortran 95 for optimal performance, whereas the outer components of the core and associated software will be developed in Java to maximize their portability. Each piece of software will be evaluated for accuracy, performance, compatibility and standards compliance. The software will be documented and released for download from the principal investigators website. A pilot project using the new software will build on the principal investigators study of stochastic gating in stellate cells of the entorhinal cortex, by evaluating the how the dendrites of stellate cells influence the functional consequences of stochastic channel gating in these neurons.
Description The primary goal of the project was to be able to efficiently simulate the activity of realistic distributions of stochastic ion channels in neurons composed of axons and dendrites with arbitrarily complex architectures. I outline below, using the measurable objectives stated in the original proposal, the development, validation and documentation of new software that achieves this goal.

1) The stochastic core. We first developed a new computational core that runs simulations of stochastic ion channel activity in model neurons that have arbitrarily complex morphological structures. The core was implemented as a compact and optimized Fortran program that we have compiled to run on Unix, PC and OS X machines. To simulate stochastic ion channel activity the core uses a new version of the tau leap algorithm that gives substantial efficiency improvements compared with the fastest previously available software for stochastic ion channel simulations. To enable easy user control we also developed an outer shell program, implemented in java and able to run on Unix, PC and OS X machines. We have called this new software the Parallel Stochastic Ion Channel Simulator (PSICS). The new software successfully passed our evaluation against the following three sets of criteria that were outlined in our original proposal:

a) Accuracy. Simulations with the new software produced results exactly as predicted from well-established theoretical principles and experimental results (e.g. see Figures 1 and 2).

b) Performance. Profiling demonstrated that the timing of the core calculation is dominated by the essential costs of the stochastic channel update procedures. Comparison of simulation times with the previous best available software indicate reductions of greater than one order of magnitude as originally anticipated. We achieved further efficiency savings by enabling simulations to be carried out in parallel when the software is run in an environment that has multiple processors available. We have successfully run the parallel version of the software on a small local network (8 processors) and on the Edinburgh Compute and Data Facility high performance cluster of servers (maximum 1456 processors).

c) Compatibility and standards compliance. The stochastic core simulates models using specifications built on the cross-platform neuroML format. The software can also directly important neuronal morphology data specified in online databases (e.g. We have carried out detailed simulations with > 15 test neurons with realistic morphology downloaded from the neuromorpho website. We have also successfully run models exported from the deterministic simulation package NEURON. We originally intended that a second modified version of the core would be operational as a plug-in to the Genesis 3 simulation environment, however unanticipated delays in Genesis 3 development that were outside of our control prevented us from undertaking this development.

2) The channel localizer tool and extension of the channelML data format. We developed a second new software tool (which we have called ICING) that deals only with the anatomical localization of ion channels on neurons with complex morphologies (Figure 3). The channel localizer tool successfully carries out the following functions: a, generate scalable 2D and 3D graphical representations of model neurons showing the distribution of ion channels as a density map or as individual points; b, transform density descriptions into model descriptions that specify the location of individual ion channels; c, transform model descriptions that specify channel location into density descriptions; d, placement of individual ion channels onto a model neuron or the use of user-defined algorithms to specify the distribution of ion channels on the membrane of a model neuron. To represent models the channel localizer tool uses data formats built on morphML and channelML specifications.

3) Distribution. We have set up a website ( for distribution and documentation of the software. Once a manuscript describing the new software has successfully completed peer review then the software and associated documentation will be made freely available to the scientific community for download from the website and the source code will be released under the General Public License (
Exploitation Route The software has been used for research by other groups and as a tool for teaching.
Sectors Digital/Communication/Information Technologies (including Software),Pharmaceuticals and Medical Biotechnology

Description The software we developed has been adopted by other research groups and has been used as a tool for teaching. Our generation of well-validated and high quality software through a contract to a commercial developer has also been used as a model for best practice in development of academic software. The commercial developer is an SME which has used the project as a model for further development of its business.
First Year Of Impact 2008
Sector Digital/Communication/Information Technologies (including Software),Education
Impact Types Economic

Title Computational models of entorhinal cortex circuits 
Description We have developed models that account for spatial firing and oscillatory activity observed in the entorhinal cortex. 
Type Of Material Computer model/algorithm 
Year Produced 2013 
Provided To Others? Yes  
Impact The model is available freely from an archive and we do not monitor its use. We're aware of the model being used by other research groups and as a tool in teaching. 
Title Models of neurons containing stochastically gated ion channels 
Description We have developed a number of models of dendritic neurons containing stochastically gated ion channels. 
Type Of Material Computer model/algorithm 
Year Produced 2010 
Provided To Others? Yes  
Impact The models were made available via an online repository. We have not monitored their use. 
Title PSICS 
Description Software for parallel simulation of stochastic ion channel gating in neurons with dendritic morphologies. 
Type Of Technology Software 
Year Produced 2008 
Open Source License? Yes  
Impact The software has been used by others for research and for teaching.