Anders Barth, Claus Seidel
All analysis was done in MATLAB. The code is hosted in a git repository.
The minimal number of states and their FRET efficiencies are determined by Gaussian fitting of frame-wise FRET efficiency histograms. We used a Gaussian mixture model as implemented in the fitgmdist
, based on an iterative Expectation-Maximization algorithm of the likelihood function.
We employ fluorescence correlation spectroscopy to analyze the time traces of donor and fluorescence signal. Three different approaches are used:
We apply the algorithm presented by (Aggarwal, 2012) [1] to identify steps in the FRET efficiency traces. The algorithm does not assume any particular kinetic scheme but estimates the optimal number of steps based on the noise of the signal. Overfitting avoided by introducing a penalty for each transition. For the analysis, we set an estimated noise of based on the distribution width obtained from the Gaussian fitting analysis ($\sigma_E$ = 0.05-0.1). An exemplary result of the step-finding is shown in Figure 1.
After the step finding, the stepwise FRET efficiency histograms were examined to identify thresholds to digitize the FRET efficiency trajectory.
[1]: T. Aggarwal, D. Materassi, R. Davison, T. Hays, M. Salapaka, Detection of Steps in Single Molecule Data. Cel. Mol. Bioeng. 5, 14–31 (2012).