Imposing Resolution on Single Channel Data Containing Sublevels

Abstract

Several classes of ion channel, including AMPA-type glutamate receptors, exhibit single channel events of different amplitudes (‘sublevels’). In some cases, sublevel behaviour might reflect fundamental principles of channel gating (e.g. ROSenmund et al., Science 1998). Such observations provide motivation for developing analysis methods that treat sublevels correctly. All idealisation methods (including threshold crossing, time course fitting and segmental k-means) have a limited temporal resolution for detecting brief events. To account for this, a fixed resolution or ‘dead time’ is normally imposed post hoc on the idealised data, on the basis of a low frequency of false event detection (‘error rate’). This error rate always varies steeply with the signal to noise ratio (SNR). In records with sublevels, SNR - and thus the error rate - varies in a discrete way from event to event. Small events have a lower SNR and thus a dramatically higher error rate. By inverting the relation between error rate and SNR, I propose a simple method to take account of the amplitude dependence of the error rate in these records, by varying the resolution in order to hold the error rate constant across the data set. Using simulated data, I show the distortions of the apparent duration of events that are introduced by imposing conservative, cavalier and variable resolutions. Although this approach complicates data display, the correct treatment of missed events in subsequent mechanism fitting should now be possible. Thus, extensions of missed event correction to data containing sublevels should consider explicitly accounting for event-dependent resolution.