ExTrack permits to assess a wide range of multi-state diffusion models. (a) Example track of a molecule transitioning between immobile and diffusive states with d₁ = 2 σ. Arrows: observed displacements; dots: actual positions of immobile (blue) and diffusive (red) molecules. (b) Consecutive observed distances of the track from a. (c) Density function of observed distances of coefficiently immobile (blue) or diffusive (red) molecules for d₁ = 2 σ. (d and e) Left: Simulated two-state (d) and three-state (e) diffusion models with diffusion length and transition rates as indicated. Right: Model parameters estimated by ExTrack (mean ± SD) assuming a two-state (d) or three-state (e) model (localization error σ, diffusion lengths d₀ and d₁, initial immobile fraction F₀, transition rates k). Dotted lines: ground truth. ExTrack settings: two-state data: two sub-steps, window length = 10; three-state data: no sub-steps, window length = 7. (f) Heatmap of the relative errors of d₁, F₀, k_u, and k_b obtained from a two-state model fit to two-state simulations as in d. Error: mean absolute relative errors from 10 replicates per condition. White lines indicate regions of <10% error for model parameters hardest to fit for ExTrack (k_u, solid), vbSPT, and anaDDA (k_b, dashed, see g). Initial parameters: d₀ = 0 µm, d₁ = 0.1 µm, σ = 0 µm, k = 0.1 Δt⁻¹, F₀ = 0.4. (g) Error on k_b of vbSPT and anaDDA (same protocol and color map as in f, 20 replicates were used for anaDDA to mitigate its variability). See Fig S2 a for errors on the other parameters.