Comparing Single Molecule Tracking and correlative approaches: an application to the datasets recently presented in Nature Methods by Chenuard et al

approaches: an application to the datasets recently presented in Nature Methods by Chenuard et al. Carmine di Rienzo and Paolo Annibale 1 NEST, Scuola Normale Superiore and Istituto Nanoscienze-CNR, Piazza San Silvestro 12, 56127 Pisa, Italy ([email protected]) 2 Center for Nanotechnology Innovation @NEST, Istituto Italiano di Tecnologia, Piazza San Silvestro 12, 56127 Pisa, Italy 3 Department of Pharmacology, University of Wuerzburg, Wuerzburg, Germany ([email protected]). Recent efforts to survey the numerous softwares available to perform single molecule tracking (SMT) highlighted a significant dependence of the outcomes on the specific method used, and the limitation encountered by most techniques to capture fast movements in a crowded environment [1]. Other approaches to identify the mode and rapidity of motion of fluorescently labeled biomolecules, that do not relay on the localization and linking of the images of isolated single molecules are, however, available [2, 3]. In particular, the molecular Mean Square Displacement (MSD) can be easily measured without identifying individual molecules positions by quantifying the enlargement in time of the Spatio-Temporal Image Correlation Function (STICS), the so called iMSD approach [2]. Alternatively, the iMSD can be applied to the spatio-temporal correlation of localization datasets, building on the improved spatial resolution obtained by the molecular localization process [3, 4]. We set to apply each of these methods to the benchmark dataset introduced by Chenouard et al. [1]. We chose the scenario where a "receptor" molecule undergoes diffusion in between intervals of directed motion, associated to "transport" along the cytoskeleton. Three exemplary levels of signal-to-noise (SNR) and particle concentrations were investigated as shown in Fig. 1. The conventional SMT analysis fails to quantitate accurately the presence of a directed motion (Fig. 1 a-d-g), due to the difficulty of correctly relinking the localizations of the fast moving molecules. This results in a sizable underestimation of the MSD, particularly at longer delays as shown by Chenuard et al. [1]. On the other hand, the iMSD (see supplementary information for further details) allows an accurate recovery of the expected MSD (Fig. 1 b-e-h). When this approach is applied to the localization datasets,