This project will develop methods of Bayesian statistical analysis for mixture distributions in studies of mechanisms of neuronal synaptic activity. Physiological analysis of synaptic mechanisms (such as short- or long-term potentiation) is based on presumptions regarding activity at individual synaptic sites, yet such individual sites can rarely be analyzed directly. Thus, to infer baseline activity and dynamic changes at individual sites from physiological recordings, statistical/neurophysiological models have been developed which include the output from multiple sites and cater for background noise. These models assume that the recorded cell output represents a random mixture distribution composed of synaptic signal 'components,' where each 'component' represents one or more synaptic sites. Bayesian mixture models and methods of analysis using (uncertain) mixtures of (uncertain numbers of) components will be explored in this context. This pilot project will aim to benchmark and validate the novel Bayesian approaches in the context of experimental data from synapses with small numbers of neural transmitter sites. Technically, research will formulate prior distributions for model parameters based on available synaptic noise data, physiological evidence, and experimental conditions; study simulations to examine operating characteristics and explore the sensitivity and robustness to prior and model specifications; derive inferences about physiological parameters; determine methods of assessment of goodness of fit; make comparisons with existing and traditional approaches; and develop computer software development for implementation of the modelling techniques. The project represents a collaborative, cross-disciplinary initiative to develop novel methods for analysis and statistical inference concerning the mechanisms governing electrochemical signal transmission at neural junctions in animal nervous systems. Interaction between the theoretical development and physiological data will play a critical role in identifying and resolving uncertainties about physiological processes and mechanisms in nervous systems, assessing site to site variability in neural signals, synaptic output, identifying critical synaptic characteristics, and assessing changes under differing experimental and physiological conditions. The project will involve substantial analysis of existing and forthcoming physiological data sets in order to provide assessments of the validity of the statistical models and the derived inferences with respect to underlying physiology. The immediate objective is to validate the new statistical approaches proposed by demonstrating their application and improvements over existing statistical analyses in synaptic response analysis. Ultimately these methods are expected to contribute to fundamental understanding of signalling mechanisms in human and other nervous systems.
