Incorporating photo-artifacts into the analysis of single molecule FRET measurements

Single molecule experiments, such as those measuring FRET, encode precise information on individual molecules free of bulk averaging. Nevertheless, because smFRET assessments employ conventional fluorescence microscopy setups, artifacts such as background photons, shot-noise, cross-talk, and fluorophore blinking are limiting factors that necessitate the use of advanced analysis methods. In our recent work we employ infinite hidden Markov models and propose a novel formulation to model and analyze smFRET measurements that contains nonparametric statistics with an explicit representation of the fluorophore photo-physics. As a result, our method combines the advantages of traditional hidden Markov models while it avoids the associated state-space size restrictions or other model selection assumptions. For this, we carefully formulate the measurement process itself accounting for the photo-physics on the individual donors and acceptors in addition to the other experimental characteristics. As a result, our framework obtains accurate estimates even when recorded intensities are excessively noisy, photo-blinked, and also of short duration. Finally, our estimates remain accurate even when fluorophore photo-states alternate at scales comparable to those of the FRET efficiency itself.