Noise reduction in single-molecule fluorescence trajectories of folding proteins

A method for noise reduction in stochastic trajectories of single molecules is described. The method generalizes a nonlinear filtration technique developed by Chung and Kennedy for single-channel recording [J. Neurosci. Meth. 40 (1991) 71] and applies it to the case of two correlated trajectories, as in a fluorescence resonance energy transfer experiment. Thus it is particularly suitable for single-molecule studies of protein folding. It is shown that the nonlinear filter facilitates the detection of various intermediate conformational states in a noisy trajectory, and can provide better estimates for the temporal positions of jumps between states and dwell times than a standard low-pass filter. It is finally suggested that the filter can also be of use for directly resolving molecular motion through the transition state region on the folding energy landscape.