A fast Markovian method for modeling channel noise in neurons.

Channel noise results from rapid transitions of protein channels from closed to open state and is generally considered as the most dominant source of electrical noise causing membrane-potential fluctuations even in the absence of synaptic inputs. The simulation of a realistic channel noise remains a source of possible error. Although the Markovian method is considered as the golden standard for appropriate description of channel noise, its computation time increasing exponentially with the number of channels, it is poorly suitable to simulate realistic features. We describe here a novel algorithm at discrete time unit for simulating ion channel noise based on Markov chains (MC). Although this new algorithm refers to a Monte-Carlo process, it only needs few random numbers whatever the number of channels involved. Our fast MC (FMC) model does not exhibit the drawbacks due to approximations based on stochastic differential equations and the values of spike jitter are comparable to those obtained with the true Markovian method. In fact, we show here, that these drawbacks can be highlighted in the approximation based on stochastic differential equation methods even for a high number of channels (standard deviation of the 5th spike is about two-fold larger than that of MCF or true Markovian method for 5000 sodium channels). The FMC model appears therefore as the most accurate method to simulate channel noise with a fast execution time that does not depend on the channel number.