Ionic Channel Current Burst Analysis by a Machine Learning Based Approach

Abstract

A new method to analyze single ionic channel current conduction is presented. It is based on an automatic classification by K-means algorithm and on the concept of information entropy. This method is used to study the conductance of multistate ion current jumps induced by tetanus toxin in planar lipid bilayers. A comparison is presented with the widely used Gaussian best fit approach, whose main drawback is the fact that it is based on the manual choice of the base line and of meaningful fragments of current signal. On the contrary, the proposed method is able to automatically process a great amount of information and to remove spurious transitions and multichannels. The number of levels and their amplitudes do not have to be known a priori. In this way the presented method is able to produce a reliable evaluation of the conductance levels and their characteristic parameters in a short time.