An approach based on fuzzy clustering and an original validation index improves single channel ionic current evaluation

A machine learning approach based on the fuzzy clustering for the study of single channel ionic current at several conductance states is presented. The procedure is able to cluster current jumps values into an optimal number of well separated parti-tions corresponding to the conductance multistate levels by the use of an original fuzzy validation index, the Partition Sum of Squares (PSS). This procedure allows to discriminate the single channel from multi channels and spikes due current in a more efficient way by the use of open channel noise. The method presented here may be applied either to step–like or burst–like single ionic current jumps, it has been used to study the burst–like single ionic current jumps generated by tetanus toxin (TeTx). The relative recorded current signal represents a typi-cal methodological case of interest presenting several conduct-ance, current multi channels and spikes.