Mixtures of self-modeling Bayesian adaptive regression splines.

摘要

Electrical stimulation of the excitatory motor nerve, which innervates the opener muscle in the crayfish walking leg, results in graded excitatory postsynaptic potentials (EPSPs). These EPSPs are composed of quantal unitary events from various synapses along the nerve terminal. Monitoring a small region of the nerve terminal, with a loose patch electrode, allows field EPSPs (fEPSPs) to measured. The discrete region of the nerve terminal is stochastic in the production of the fEPSPs in relation to nerve terminal depolarization. The variation in the shape of the fEPSPs can occur by various biological mechanisms, so it is important to determine a means in indexing the shapes. The time constraints of characterizing the shapes of these events can only be reasonable done in large numbers by automated procedures.; The fEPSPs appear to follow a self-modeling regression, with affine variation in both the voltage and time axes. The underlying firing function is estimated with Bayesian Adaptive Regression Splines (BARS, DiMatteo et al. 2001), and a mixture model is implemented to determine which current traces are firings and non-firings. For the firings, point estimates and credible intervals for the coefficients of the affine transformations are calculated.; Keywords. BARS, Mixtures, Regressioh, Splines, EPSPs