Skip to content
FrancoisSimon edited this page Jun 27, 2022 · 6 revisions


ExTrack is a method to determine kinetics of particles able to transition between different diffusion states. It can assess diffusion coefficients, transition rates, localization error as well as annotating the probability for any track to be in each state for every time points. By using all spatio-temporal information contained in tracks. To do so, ExTrack computes the track probability by considering all possible sequences of hidden states and actual particles positions and performs integration over these metrics to obtain analytical expression of the track probability given a set of model parameters. This allows us to first find the most likely set of parameters of a given data set by Maximum Likelihood Estimate (MLE). Then, given a set of parameters it allows to evaluate the probability of tracks to be in each state at every time point. It can also produce histograms of durations in each state to highlight no markovian transition kinetics. Eventually, it can be used to refine the localization precision of tracks by considering the most likely positions which is especially efficient when the particle do not move. We have shown this method to be accurate even with low diffusion compared to localization error which makes it especially relevant in the hard cases where the user has to differentiate immobile particles from slowly diffusive ones. It is also able to assess a wide range of transition rates and shown to be robust when particle motion deviates from initial assumptions.


  • Multi-plateform: ExTrack is available both as a python package and as a tool integrated to the Trackmate plugin in Fiji. v.
  • Fast computation: The introduction of a user-defined window where all possible sequences of states are considered allow ExTrack to do accurate parameter estimates with fast computation time. Outside the window, possible states are averaged according to their respective probabilities.
  • Multi-state analyses: ExTrack works in principle for any number of states. However, computation time increases with number of states and a trade-off between number of state and window length has to be found. See for more details.
  • Treat localization error as a parameter or an input: Localization error can be a unique or multiple parameters for each spatial dimensions. It can also be specified for each peak based on a prior method. See for more information.
  • Multi-task: Can perform parameter fitting at the population, single-molecule probabilistic state annotation, position-refinement when particles are immobile (or moving slowly) and state duration histogram to detect deviation from time independent markovian transitions.
  • Gives correct parameter estimations for a wide range of parameters.
  • Robust to deviations from initial model assumptions.

How to cite ExTrack

to be seen when available in bioaRxiv