Get Started

ExTrack is available in Fiji from the plugin TrackMate. For more information on TrackMate, see https://imagej.net/plugins/trackmate/ . In the current Fiji implementation, ExTrack is limited to perform parameter fitting and state annotation for 2-state models but also profits from TrackMate for a nice display of probabilistic state annotations of tracks on top of the corresponding movie.

Open your movie of interest with Fiji. Choose Plugins > Tracking > Trackmate.

You should obtain:

Choose Next.

Select detector (LoG detector works fine to me). Then Next.

Estimated object diameter should be around 0.5 um in case of single-molecule tracking (to be modified depending on the needs). Select the adequate Quality Threshold depending on the pixel intensities in the movie (click on preview to check for the quality of the detection). Do not forget to use sub-pixel localization by checking the associated box as below:

Click on Next 4 times (thresholds on spots can be performed if necessary). Choose Simple LAP Tracker and adequate numbers:

Choose Next twice. Then thresholds can be put. Add a threshold on the minimum track length of 3 to 5 (a threshold on the maximal track length of 80 may also be required). Then Next and Next and Next.

Select an action > Compute ExTrack probabilities and Execute. An ExTrack window will open in addition to the TrackMate window:

In Advanced, you can select the number of sub-steps and the frame/window length of the method. A short frame length allow faster convergence at little extend of accuracy of parameters as long as diffusion is high enough. number of sub-steps should be 1 by default (no sub-steps) but can be set to 2 if transition rate per frame are more than 0.4. In a first time, we advise to keep Nb sub-steps to 1 and N frames to 6. See https://github.com/vanTeeffelenLab/ExTrack/wiki/Trade-off-for-fast-computing for more details.

In Maximum likelihood estimate, One can perform model parameter fits by selecting start estimation. If NaN appears it is most likely due to overflow (exponential of a too high number), one can try to only select short tracks to avoid this issue (e.g. tracks of 6 positions). Once the experimenters have an idea of the correct set of parameters, they can specify them in the manual input section and potentially restart a fit which should work better.

Parameters are:

• Localization error: The standard deviation of the Gaussian localization uncertainty function.
• Diffusion lengths for both states: The standard deviation of the Gaussian jump distances between sub-subsequent physical positions.
• Fraction in diffusive state: The initial fraction in diffusive state for first positions of each track F1 which also corresponds to the global fraction in diffusive state as this implementation assumes steady-state (equilibrium of fractions).
• Probability of unbinding: probability of unbinding pu which can be turned to transition rate ru per step by computing ru = - log(1 - pu) the binding rate can then be computed as such: rb = ru * (1 - F1) / F1

The user can then click on compute probabilities. State probabilities can then be displayed. In the TrackMate window, click twice on the <== box to come back to the display options panel. In the box color tracks by: select probability diffusive

Show entire tracks can also be replaced by Show tracks backward in time with fade range of for instance 60 to get a better sens of the temporality of tracks.