Erik Lundin: Simulation, analysis and visualization of single-molecule tracking data

Abstract

Single-molecule tracking allows researchers to observe individual molecules in cells. By labelling protein and RNA molecules with fluorophores, their movement in the cell can be followed using a fluorescence microscope. The molecules are detected as individual spots and their positions over time connected to trajectories. The trajectories are analysed, e.g., by fitting a hidden Markov model to infer diffusional states and transitions between them. This kind of analysis is, however, usually limited in resolution power and accuracy, introducing unknown biases. To evaluate and correct for these biases, analysis of simulated microscopy movies generated from a known reaction-diffusion model can help. This thesis focuses on the analysis and visualization of single-molecule tracking data and the fitted models, as well as the use of simulations to improve model interpretation.

Paper I investigated mRNA translation in live cells by tracking of ribosomal subunits. The duration of translation events was measured and the results suggested that re-initiation on a downstream open reading frame on the same mRNA is rare for 30S subunits, while it might be relatively frequent for 70S ribosomes.

Paper II investigated the signal recognition particle (SRP) system which targets proteins for insertion into the membrane. The 2D tracking did not allow direct observation of individual membrane interactions. However, by analysing the spatial distributions it was possible to estimate the occupancy and dwell time of the membrane association of SRP. To validate the model, simulated microscopy movies were generated based on a reaction-diffusion model and analysed the same way as experimental movies. To obtain a plausible ground truth model, the parameters of the reaction-diffusion model were iteratively adjusted to align the analysed simulation results with the experimental results.

Paper III explored single-molecule tracking using double helix phase masks that enabled localization in 3D. Here we investigated membrane association of ribosomes translating inner-membrane proteins. To evaluate the method, a simulated membrane-cytosol reaction diffusion model was used. By applying a circular error measurement, individual membrane-ribosome binding and release events could be detected and quantified. The membrane interaction model was combined with a diffusional state model to identify the biological states:  50S, cytosolic 70S and membrane 70S.