Hidden Markov analysis of improved bandwidth mechanosensitive ion channel data.

The gating behaviour of a single ion channel can be described by hidden Markov models (HMMs), forming the basis for statistical analysis of patch clamp data. Extensive improved bandwidth (25 kHz, 50 kHz) data from the mechanosensitive channel of large conductance in Escherichia coli  were analysed using HMMs, and HMMs with a moving average adjustment for filtering. The aim was to determine the number of levels, and mean current, mean dwell time and proportion of time at each level. Parameter estimates for HMMs with a moving average adjustment for low-pass filtering were obtained using an expectation-maximisation algorithm that depends on a generalisation of Baum's forward-backward algorithm. This results in a simpler algorithm than those based on meta-states and a much smaller parameter space; hence, the computational load is substantially reduced. In addition, this algorithm maximises the actual log-likelihood rather than that for a related meta-state process. Comprehensive data analyses and comparisons across all our data sets have consistently shown five subconducting levels in addition to the fully open and closed levels for this channel.