Level detection in ion channel records via idealization by statistical filtering and likelihood optimization

Abstract

A parameter–free method is presented for the level detection in ion channel records via recovery of step wise current changes. No assumptions about ion channel mechanism are made. The primary detection of the transitions is made by statistical filtering the data using the Student'st–test. The event currents are calculated as the average value of the current between two adjacent transitions. An optimal ideal trace is found by maximization of a likelihood function. The distribution of event currents recovered from the raw data is then analysed, again by using the Student'st–test, for their grouping into separate statistical ensembles, defining current levels. The method is subjected to rigorous test using simulated data, and is compared with several other methods. It produces the levels of channel current, their noise amplitudes and distributions of dwell times, the desired information for constructing the channel mechanism.