To the Editor:
Single-molecule fluorescence resonance energy transfer (smFRET) with total-internal-reflection fluorescence (TIRF) microscopy is a powerful tool for studying the structure and conformational dynamics of biomolecules1,2,3. By reporting on nanoscale distances within single biomolecules, this technique provides the means to uncover population heterogeneity that is normally hidden behind an ensemble mean as well as to monitor the dynamic equilibria of unsynchronized samples4. smFRET additionally holds the potential to determine the structure of dynamic molecules in solution at low molecular concentration5. However, the processing, analysis and interpretation of the large amounts of raw image data in smFRET experiments can be complex and time consuming.
Here we present iSMS: an interactive toolkit for the comprehensive analysis of smFRET TIRF-microscopy data (Fig. 1a and Supplementary Software). iSMS integrates and automates common procedures in smFRET data analysis: molecule localization, intensity-trace integration, quantitative FRET determination, FRET distribution analysis, molecule subpopulation analysis and transition state dynamics analysis (Supplementary Notes 1–6 and Supplementary Figs. 1,2,3,4,5,6,7,8,9,10). The software provides tools for interacting visually with the data and enables the production of publication-friendly figures. The strengths of iSMS lie in its immediate usability, flexibility and analysis speed and in the several built-in tools for extracting, evaluating and clustering all molecules identified in the raw image data. The entire program can be operated without any prior programming skills and allows interactive sessions of raw and processed data to be easily shared, explored and evaluated by peers.
iSMS processes image data almost 20 times faster than current software standards (Supplementary Note 7, Supplementary Table 1 and Supplementary Fig. 11) and is used both in parallel to and after data acquisition. Parallel computing is supported for time-consuming routines, and tools for monitoring and allocating memory are built into the interface. The toolkit can be applied to analyze raw video data from most experimental setups and major microscopy and camera suppliers (Supplementary Note 1). Several robust algorithms help automate the data analysis via interactive tools (Supplementary Notes 2–6). Automation includes emission-channel registration, drift correction, peak localization, FRET-pair identification, intensity integration, FRET time-trace determination (Fig. 1b), detection of bleaching, determination of correction factors, molecule filtering and grouping (Fig. 1c), dwell-time analysis and distribution analysis (Fig. 1d). The program is distributed in compiled and open-source versions at http://isms.au.dk.
Roy, R., Hohng, S. & Ha, T. Nat. Methods 5, 507–516 (2008).
Selvin, P.R. & Ha, T. Single Molecule Techniques: A Laboratory Manual (CSH Press, 2008).
Hohlbein, J., Craggs, T.D. & Cordes, T. Chem. Soc. Rev. 43, 1156–1171 (2014).
Hildebrandt, L.L. et al. J. Am. Chem. Soc. 136, 8957–8962 (2014).
Preus, S. & Wilhelmsson, L.M. ChemBioChem 13, 1990–2001 (2012).
Bronson, J.E., Fei, J., Hofman, J.M., Gonzalez, R.L. Jr. & Wiggins, C.H. Biophys. J. 97, 3196–3205 (2009).
This work was supported by grants from the Danish National Research Foundation to the CDNA Center (grant number DNRF81), from the Danish Council for Independent Research's carrier program Sapere Aude and from Aarhus University Research Foundation.
The authors declare no competing financial interests.
Integrated supplementary information
The figure shows an image registration between donor (green) and acceptor (red) emission channels. (a) The algorithm starts with an initial guess and (b) returns the positions of the aligned ROIs. The black inserts are overlay images in which the green color channel is the donor ROI and the red color channel is the acceptor ROI image data. The raw image frame is of size 512x512 pixels in this example.
The peakfinder algorithm localizes donor (green) and acceptor peaks (red) and identifies FRET-pairs (yellow) based on the relative distance in between the found peaks.
(a) Molecule image of the donor, acceptor and direct acceptor emission frame, respectively. (b) A local mask is used to integrate the intensity (colored trace, white mask) and calculate the background (black trace, grey mask). (c) The intensity after bleaching is used to calculate the background.
Supplementary Figure 4 Detection and correction of horizontal sample drift during image acquisition using iSMS.
(a) Detected drift direction (left) and magnitude (right). (b) Effect of drift correction on a molecule intensity trace. The example shown is for a real data set and real molecule time trace, but the drift was added synthetically for demonstration purposes.
These correction factors are calculated using molecules in which the acceptor bleaches before the donor. In (a) the gamma factor correction is determined from the average fluorescence intensities D1, D2, A1 and A2. The lower graph shows the FRET efficiency. In (b) the lower graph shows the donor leakage correction factor trace in which the red line is the average value. The highlighted regions depict the time-intervals used for calculating the correction factors.
Supplementary Figure 6 Calculation of the direct acceptor correction factor in measurements using alternating laser excitation (ALEX).
The direct acceptor correction factor is calculated using molecules in which the donor bleaches before the acceptor. The lower graph shows the correction factor trace in which the red line is the average value. The highlighted regions depict the time-intervals used for calculating the correction factor.
(a) Detection of donor bleaching using the sum of donor and acceptor fluorescence time traces obtained after donor excitation; D+A sum trace (bottom). (b) Detection of acceptor bleaching using the acceptor fluorescence time trace after direct excitation of the acceptor, direct AA reference trace (bottom panel).
(a) Fitting the ideal path of a single FRET trace using Hidden Markov Modelling (HMM). Three states (I, II and III) are identified by the algorithm and the time intervals the molecule spends in each of the three states. (b) FRET histogram of isolated states. This corresponds to the FRET histogram within the time intervals defined by the state time in the idealized trace. (c) Dwell times of multiple molecules. The scatter plot provides information on all fitted states while the insert shows a binned histogram of an isolated FRET-interval defined by the red box. The histogram is a measure of the decay kinetics of that particular FRET state. (d) Transition density plot: a binned histogram of the number of transition events occurring from one FRET level to another.
In this example, three different molecule behaviours are observed within the same sample: 1) molecules trapped in a low FRET state (an example in lower left), 2) molecules trapped in a high FRET state (an example in upper left) and 3) molecules dynamically alternating in between at least two different FRET states (upper right). The black traces are backgrounds, green traces are raw donor intensities, red traces are raw acceptor intensities upon donor and acceptor excitation, respectively, and the blue traces are stoichiometry and FRET efficiency respectively. The population histogram (lower right) quantifies the subpopulation distributions visually. In the examples, donor bleaching time is highlighted in green and acceptor bleaching is highlighted in red.
(a) Prediction of number of Gaussian components in the 2D FRET Stoichiometry data by variational Bayesian inference. Each predicted 2D component has its own color. (b) Fit of Gaussian mixture components to each of the 1D data sets: FRET (top, three components) and stoichiometry (right, two components).
(a,b) Screenshots of the (a) TwoTone and (b) iSMS main windows. (c) Comparison of computational processing times for extraction of FRET traces from the same raw movie data set. In total, iSMS performed 17 times faster than TwoTone for this data set. TwoTone found 185 molecules and iSMS found 241 molecules. TwoTone software: Holden, S. J. et al. Biophysical Journal 99, 3102–3111 (2010)
Supplementary Figures 1–11, Supplementary Table 1 and Supplementary Notes 1–7 (PDF 14991 kb)
iSMS software source code running in MATLAB (ZIP 11192 kb)
Compiled version of the iSMS software for Windows 32-bit (ZIP 32412 kb)
Compiled version of the iSMS software for Windows 64-bit (ZIP 33083 kb)
About this article
Cite this article
Preus, S., Noer, S., Hildebrandt, L. et al. iSMS: single-molecule FRET microscopy software. Nat Methods 12, 593–594 (2015). https://doi.org/10.1038/nmeth.3435
Molecular Cell (2020)
The FEBS Journal (2020)
DeepFRET, a software for rapid and automated single-molecule FRET data classification using deep learning
Polymerase μ in non-homologous DNA end joining: importance of the order of arrival at a double-strand break in a purified system
Nucleic Acids Research (2020)
DNA origami-based single-molecule force spectroscopy elucidates RNA Polymerase III pre-initiation complex stability
Nature Communications (2020)