Representing Idealizations of Ion Channel Records

J. Alex Stark, Rafael Rosales, William J. Fitzgerald and Stephen B. Hladky

43rd Meeting of the Biophysical Society, February 1999. Abstract in Biophysical Journal, 76(1):A199 (1999)

Bayesian statistical idealization of ion channel records produces a distribution of possibilities rather than a single ideal trace. Summarizing this (posterior) distribution and displaying informative realizations requires new procedures. Here we address several difficulties and present two methods for summarizing the output. Conductance versus time is displayed as the mean together with skewness markers that indicate the spread in either direction. Examples show that these markers provide a better indication of the possible level positions than mean +/-standard deviation. When a histogram with logarithmic bins is constructed for quantized durations, there are severe irregularities in the counts per bin, with many empty bins for short durations. Unevenness in the mapping can extend to durations greater than 100 sample intervals. We show how to use this histogram to construct an estimate for the probability density function (pdf) with a logarithmic time axis. With an algorithm such as ``ALL'' which does not assume a small number of conductance levels it is necessary to display a two-dimensional (2D) pdf versus log(duration) and conductance rather than a family of 1D pdf's for each of the levels. The 2D pdf corresponding to a single level with an exponential distribution of durations has a spread over conductance which increases as the duration decreases. The peak of the 2D pdf occurs at 1.5x the time constant for the exponential.