Book Description

Ribonucleic acid (RNA) is an essential biomolecule in the domain of life. As a mediator between the genetic information contained in deoxyribonucleic acid (DNA) and the cell, its regulation has a major impact on the homeostasis of living organisms. The discovery and identification of chemical modifications at nucleotide level in RNA has made it possible to study a new layer of regulation of these molecules. In view of the importance of these chemical alterations within the biology domain, a specific term for the study of these modifications has been coined: epitranscriptomic. More than 170 modifications have now been listed, and their analysis has led to the discovery of numerous links between these alterations and various diseases such as viral infections, cancers and neurodegenerative diseases. Epitranscriptomic therefore holds out the hope of developing innovative treatments for yet incurable diseases. However, of the large number of modifications identified to date, few methods exist that can locate them precisely in RNA sequences. In order to overcome these limitations, the work presented here concerns the development and optimisation of methods for detecting modifications through high-throughput sequencing. More specifically, the focus of this work is on the computational processing of sequencing data in order to perfect the detection of these modified nucleotides. As a first step, an application of a method for detecting 2'-O-methylations via high-throughput sequencing entitled RiboMethSeq was carried out on tRNA samples from Escherichia coli and Saccharomyces cerevisiae with the aim of studying the role of these modifications in innate immunity. This study uncovered interesting details about the regulation of immunity, but also revealed detection limitations on the part of RiboMethSeq. These limitations are mainly due to the non-optimal parameterisation of the various data processing stages, but also to the lack of transfer RNA (tRNA) sequence references suitable for epitranscriptomic studies. These limitations have been specifically addressed and a thorough optimisation of these two concepts has been implemented, enabling a more in-depth analysis of the links between 2'-O-methylations and innate immunity in a second study. Finally, with the aim of pushing back the limits of 2'-O-methylation detection, the potential of deep learning algorithms for detecting modifications is explored using RiboMethSeq ribosomal RNA (rRNA) data as a training set for Random Forest algorithm. Secondly, development of alternative methods for the detection of two modifications, N6-methlyadenosine (m6A) and pseudouridine (Psi) is carried out. Existing detection methods for these modifications have limitations due to the nature of the approach chosen. In order to fill these gaps, two alternatives processes based on the induction of chemical signatures emitted by these modifications were proposed and then applied to biological samples, proving their robustness of detection as well as their quality of quantification of these biomolecules of interest. Finally, this work concludes with the simultaneous use of three detection methods - RiboMethSeq, AlkAnilineSeq and BisulfiteSeq - on the same set of brain cell samples in the context of studies on neurodegenerative diseases. The combination of these methods enabled five human tRNA modifications to be mapped, and allowing their respective quantification according to the nature of the cell tissue and stress conditions. This work has enabled us to confirm in greater detail observations already seen in the literature, but also to highlight a still little-studied modification, Dihydrouridine, as a potential determining factor in tRNA fragmentation.