New Analysis Tools to Characterize the Diffusive Behavior of Single Protein Trajectories in Live Cells

Abstract: Recent advances in live cell imaging have now made it possible to non-invasively monitor the motions of individual proteins inside live cells. Therefore, over the past few years, interest in improving the quality of the diffusion analysis for single particle trajectories has grown rapidly. However, the photoinstability of the fluorescent probes used in these approaches results in a premature truncation of the observed protein trajectories. Due to the limited statistics provided by these short trajectories, traditional mean squared displacement analysis is unreliable, because these approaches generally fail to properly handle noise sources in a statistically accurate manner, inevitably yielding suboptimal results. Here, we introduce new analysis tools to characterize different aspects of the underlying protein dynamics: the number of distinct diffusive states (perturbation expectation-maximization for short-time diffusive states and variational Bayes expectation-maximization for macroscopic diffusive states); the diffusive properties of each state (maximum likelihood estimator for non-normal modes of diffusion); and classification of individual particle trajectories to a respective diffusive state (likelihood Bayes). We test the performance of these new analysis tools on various sets of simulated particle tracks subject to static and dynamic localization noise. We then demonstrate the applicability of these analysis tools on single protein trajectories of Rho GTPase, an integral regulator of cytoskeletal dynamics and cellular homeostasis, in live cells acquired via single particle tracking photo-activated localization microscopy.