Classifying directed and diffusive transport in short, noisy single-molecule trajectories with wMSD
Fluorescence imaging in combination with single-particle tracking analysis has emerged as a powerful tool to study and characterize the motion of proteins moving in biological media. One of the main challenges in this approach is to reliably distinguish between directed and diffusive transport, especially for short and often noisy trajectories showing distinct, time- and place-dependent modes of motility. In this contribution, we present a windowed Mean-Square Displacements classifier (wMSDc) that is able to reliably (i) identify periods of diffusive and directed transport, (ii) extract position-dependent diffusion coefficients and velocities, and (iii) identify the location of switches in direction or motility mode in short (< 50 time points) and noisy single-molecule trajectories. We compare the performance of this approach to a Hidden Markov Model (HMM) method and a Moment Scaling Spectrum based method (DC-MSS) previously published and show that, in most cases, its performance is superior. We present a wide range of applications: from the movement of whole organisms and cells to protein-DNA interactions in vitro and motor-protein dynamics in vivo. Statement of SignificanceExtracting quantitative parameters from single-particle trajectories is a challenging task, especially in biological samples, which often show a high degree of heterogeneity. Trajectories can reveal switches between different types of motion; directed, diffusive and sub-diffusive motion. Usually, the length of these trajectories and their localization precision are limited by the experimental conditions. Here, we present a novel approach to analyse single molecule trajectories, windowed Mean-Square Displacement classifier (wMSDc) to reliably distinguish between directed and diffusive transport in the short trajectories with a finite precision of localization and integration time typically obtained when imaging single fluorescent proteins in living cells or organisms. We show that, using simulated and a wide range of experimental trajectories, wMSDc is a reliable method to extract motility parameters such as diffusion coefficient and velocity.