| The present work extends the application of Markov models from single channel data analysis to other single molecules and to ensembles of molecules. For ion channels, an existing maximum likelihood method for estimation of rate constants was modified to work with non stationary data, recorded under arbitrary stimulation protocols. The method (MIP) requires that the Markov signal be extracted from the raw data (idealization).; For data contaminated with baseline artifacts, the baseline fluctuations are modeled with a linear Gaussian model (Kalman filter). Two maximum likelihood methods were designed that simultaneously idealize data and correct the baseline. The procedures are based on coupling the Kalman filter with either the Viterbi algorithm or the Forward-Backward procedure. The resulting methods are optimal for stochastic baseline fluctuations but work well with deterministic drift or sinusoidal interference. A method for calculating the transition rate matrix from the transition probability matrix, with linear constraints imposed on rate constants, is presented.; Recording the position of molecular motors, such as kinesin, in single molecule experiments, produces data with a staircase appearance. The kinetics of the molecular motor is modeled with a truncated hidden Markov model, reflecting the identical chemistry of each step. The signal is extracted (idealized) from the staircase data with a modified version of the baseline correction algorithm. A modified version of the MIP algorithm estimates rate constants from the idealized staircase data.; Hidden Markov models were applied to the kinetic analysis of ensemble data, such as obtained in whole-cell recordings. A method for direct estimation of rate constants was developed, that simultaneously estimates the rate constants and the number of active channels in the record. |
