Abstract

Multi-electrode arrays (MEAs) provide simulateneous extra-cellular recordings of the spiking activity of several neurons cultured in-vitro over long periods of time. However, these data are typically contaminated by various sources of noise such as spike sorting errors. The availability of these noisy recordings provides the opportunity for developing statistical models of the neurons’ firing patterns to estimate their functional characteristics and to predict their spiking behaviour as a function external stimulation.

In this seminar I will present a novel hierarchical dynamic Bayesian network model describing the spiking patterns of neuronal ensembles over time. The parameters characterizing the discrete-time spiking process, the unknown structure of the functional connections among the analysed neurons and its dependence on their spatial arrangement are introduced at separate model stages. Posterior estimates for all model parameters and predictions for future spiking states are computed via the Gibbs sampler using a shrinkage prior. The adequacy of the model is investigated by plotting the raw residuals and by applying the time-rescaling theorem. I will also illustrate the analysis of a set of experimental MEA recordings showing that one neuron has a pivotal role for the initiation and persistence of the network activity and that the estimated network structure significantly depends on the spatial arrangement of the neurons.

This is joint work with Mathisca de Gunst (Vrije Universiteit, Amsterdam) and with Jaap van Pelt (Netherlands Institute for Brain Research, Amsterdam).