|M.Sc Student||Granik Naor|
|Subject||Characterization of Intracellular Transcription|
and Diffusion Kinetics
|Department||Department of Biomedical Engineering||Supervisors||Dr. Yoav Shechtman|
|Professor Roee Amit|
|Full Thesis text|
The cellular cytoplasm is the environment in which all intracellular reactions take place. Its physical and chemical properties have a strong influence on a multitude of functions, such as signaling, transport, protein folding etc. Here, we aim to shed light on two different intracellular dynamic processes which have gained increased attention in recent years, owing to technological improvements in microscopy techniques: The formation of nuclear speckles, and the occurrence of anomalous diffusion.
First, nuclear speckles are membrane-less protein-rich bodies built around a long-non-coding RNA (lncRNA) scaffold. As a means of studying the dynamics of such a formation, we investigate the dynamics of synthetic speckles in bacteria by encoding two types of synthetic lncRNAs (slncRNA), which form the basis of the bacterial speckle. The slncRNAs incorporate RNA-binding phage-coat-protein (RBP) binding sites downstream from a pT7 promoter. For both slncRNAs studied, fluorescent speckles containing dozens of RBP-bound slncRNA molecules form in cell poles. Fluorescence measurement over time reveals both positive and negative changes in intensity spaced by exponentially distributed periods of non-classified activity. We identify positive changes with transcriptional bursts, and term the negative, fluorescence degradation bursts. The data indicates that negative bursts correspond to shedding of multiple slncRNAs back to cytoplasm.
Second, diffusion plays a critical role in many biological processes in the cell. Direct observation of molecular movement by single-particle-tracking experiments has contributed to a growing body of evidence that many cellular systems do not exhibit classical Brownian motion, but rather anomalous diffusion. Characterization of the physical process underlying anomalous diffusion remains a challenging problem due to the fact that commonly used tools for distinguishing between these processes are based on asymptotic behavior, which is experimentally inaccessible in most cases. Additionally, an accurate analysis of the diffusion model requires the calculation of many observables since different transport modes can result in the same diffusion power-law α, which is typically obtained from the mean squared displacements (MSD). We opted to use deep learning to infer the underlying process resulting in anomalous diffusion. We implemented a neural network to classify single-particle trajectories by diffusion type, separating between Brownian motion, fractional Brownian motion (FBM) and Continuous Time Random Walk (CTRW). We demonstrate the applicability of our network architecture for estimating the Hurst exponent for FBM and the diffusion coefficient for Brownian motion on both simulated and experimental data. We show these networks achieve better accuracy than time-averaged MSD analysis on simulated trajectories while requiring fewer time-steps. Furthermore, on experimental data, both network and ensemble MSD analysis converge to similar values, with the net requiring only half the number of trajectories required for ensemble MSD to achieve the same confidence interval.