Statistical Modeling and Computational Analysis of Ribosome Profiling for the Dissection of Translational Control

Abstract

mRNA translation is one of the most complex molecular processes that has been developed during the evolution of the cell. It synthesizes proteins that are the building blocks and workhorses of the cells. Translation consists of initiation, elongation and termination. These three steps require hundreds of molecules to participate in a concerted manner. Translational control regulates protein levels in response to intra- and extra-cellular environmental changes. Regulation at the translational level is important in situations where transcription regulation alone cannot satisfy the emergent needs of the cell or when local control over protein abundance is required. Translational control plays an essential role in maintaining cell homeostasis, physiology and modulating cell growth. Dysregulation of mRNA translation or aberrant function of translation machinery can lead to a variety of diseases including metabolic disorders and cancer. Thus, elucidating the mechanisms of translational control is key for understanding how diseases develop.

High-throughput sequencing technologies are widely used to determine and quantify DNA and RNA molecules on a large scale, which has remarkably facilitated our understanding of many biological functions in a system-wide manner. An extension of this technology, ribosome profiling, even allows characterization of ribosome-occupied mRNA fragments. Ribosome profiling, thus, provides an opportunity to globally monitor the translation in vivo and study the mechanisms of translational control.

My thesis consists of two main parts: 1) The first part focusses on development of RiboDiff, a statistical framework and computational tool for detecting genes under differential translational regulation. RiboDiff fits quantitative ribosome profiling and RNA-Seq measurements with a negative binomial based generalized linear model. Subsequently, a statistical test is performed to identify genes under differential translational control between measurements in two conditions. Our experiment demonstrates RiboDiff outperformed state-of-the-art existing approaches. 2) The second part establishes a computational pipeline for analyzing ribosome profiling data. Using this pipeline, we studied the translational regulation in leukemia and other conditions. We identified mRNAs that presented distinct translation efficiency and ribosome footprint density in a condition that the cells were treated with a chemical compound Silvestrol. Further analysis of these mRNAs revealed that the guanine quartet (GCC)4 sequence pattern, which forms a G-quadruplex structure, is enriched in the 5' UTR of 280 mRNAs with down-regulated translation. Experimental validations supported our findings and confirmed that the G-quadruplex is an RNA element that represses the translation initiation activity. Applications of the computational approach on other translational researches illustrate the versatility of the proposed analysis methodology.