A distribution-based method to resolve single-molecule Forster resonance energy transfer observations.

We introduce a new approach to analyze single-molecule Forster resonance energy transfer (FRET) data. The method recognizes that FRET efficiencies assumed by traditional ensemble methods are unobservable for single molecules. We propose instead a method to predict distributions of FRET parameters obtained directly from the data. Distributions of FRET rates, given the data, are precisely defined using Bayesian methods and increase the information derived from the data. Benchmark comparisons find that the response time of the new method outperforms traditional methods of averaging. Our approach makes no assumption about the number or distribution of underlying FRET states. The new method also yields information about joint parameter distributions going beyond the standard framework of FRET analysis. For example, the running distribution of FRET means contains more information than any conceivable single measure of FRET efficiency. The method is tested against simulated data and then applied to a pilot-study sample of calmodulin molecules immobilized in lipid vesicles, revealing evidence for multiple dynamical states.