Realistic molecular simulation of neuronal signalling systems

Date: 1 June 2016

Venue: UAntwerp, Campus Drie Eiken, Building R, Auditorium R1 - Universiteitsplein 1 - 2610 Wilrijk (Antwerp)

Time: 4:00 PM - 6:00 PM

PhD candidate: Iain Hepburn

Principal investigator: Erik De Schutter

Short description: PhD defence Iain Hepburn - Department of biomedical sciences


The mammalian brain is a hugely complex machine that operates over a wide range of spatial and temporal scales, calling for many different computational techniques to realise the full potential of advancing our understanding of brain function through computational power. Neurons possess a rich subcellular molecular soup that carries out many important functions on the small spatial scale, such as controlling the strength of synaptic contacts between cells. This molecular world is often grossly simplified or ignored in large network simulations where the electrical behaviour of neurons is computed under various levels of abstraction from the real system. There are, however, occasions where simple modelling of molecular systems misses key features that can affect the computations on the larger spatial scale: one example is the intrinsic noise that arises from probabilistic molecular interactions that can affect the reliability of the electrical signal. There are, therefore, many challenges to be faced to ultimately understand how neuronal molecular systems couple to electrical signalling and impact brain function.

I devoted my doctoral studies to developing computational methods that are capable of simulating subcellular neuronal molecular systems to a high level of detail, with a strong emphasis both on accuracy and efficiency. Close morphological representation of the real cells is achieved with tetrahedral mesh spatial discretization, and realistic molecular behaviour is captured by stochastic reaction-diffusion. In many cases the stochastic reaction-diffusion behaviour differs significantly from the simplified non-spatial, deterministic modelling that is often applied to model such systems. The extra detail comes at a considerable computational cost, but exact stochastic methods can be closely approximated by a parallel algorithm that is found to speed up models by orders of magnitude in a Message Passing Interface implementation.

An important step towards integrating the simulations of high molecular detail with the electrical behaviour of the cell was to develop a novel algorithm that computes the electrical potential on the same tetrahedral mesh as for the reaction-diffusion simulation, allowing close coupling between the two systems. These techniques were applied to a model of Purkinje cell dendritic calcium burst in which we observed many interesting features that are not captured with more simplified modelling approaches, both in the electrical and molecular signal. Therefore, I conclude that realistic molecular simulation will play a vital role in advancing our understanding of neuronal systems and brain function.