A Multiple Hypothesis based Particle Tracking Method with Application to Clathrin Mediated Endocytosis Analysis

The development of advanced fluorescence microscopy techniques as well as genetically encoded tags such as Green Fluorescent Protein (GFP) has radically improved our ability to observe individual particles composed of proteins and protein complexes in live cells, and to understand molecular mechanisms of the cellular and subcellular processes by analyzing the dynamics of these particles. In particular, we are studying clathrin mediated endocytosis (CME) which is a fundamental subcellular process and has house-keeping functions in all types of cells. CME is a major route for synaptic vesicle recycling at neuronal synapses, a process essential for synaptic transmission, and dysfunction of the process is implicated in several neurological and psychiatric diseases such as Alzheimer disease. Abnormal function of CME can also be involved in cancer and diabetes. To quantitatively study molecular mechanisms of this process, it is critical to be able to accurately track the involved particles over time. However, since the image datasets from an experiment typically have several thousand images, manual processing is very time-consuming, and therefore, an automatic tracking method is desired.;We have developed a multiple hypothesis based particle tracking method for clathrin mediated endocytosis (CME) analysis using total internal reflection fluorescence (TIRF) microscopy. The framework of the tracking method is an extension of the classical multiple hypothesis tracking (MHT). In the extended MHT framework, particle tracking becomes evaluating two types of hypotheses. The first one is the trajectory-related hypothesis, to test whether a recovered trajectory is correct, and the second one is the observation-related hypothesis, to test whether an observation from each image belongs to a real particle. Here, an observation refers to a detected particle and its estimated feature vector. Therefore, candidates of trajectories and observations are generated and evaluated, and the optimal subset of the candidates are selected as the final result. In the original MHT framework, only limited trajectory candidates are generated and no observation candidates are considered, resulting in a relatively low accuracy. The extensions are designed to meet the challenges imposed by high particle densities in the images: particles seem to merge and then split when the distances between them are close to the TIRF resolution limit, which may lead to wrong observations; new particles appear near the locations where matured particles disappear, which may lead to wrong trajectories made from several independent trajectories. To track the particles in CME, we have studied their properties and designed state models to describe the random and constrained motion.;The proposed method is validated in simulation under different scenarios. The results show that, compared to the MHT based methods, the proposed method achieves higher accuracies, which is attributed to the particle state models and the extended MHT framework. The proposed method is also applied to real image data to study the effect of the molecular compound, methyl-beta-cyclodextrin (MbCD) in CME by analyzing the images from the control group and the treatment group, and the significant difference in particle lifetime between the two groups is identified.