Back to search

University of Thessaly (UTH)

Υπολογιστική ανάλυση των λειτουργιών των μη κωδικών μεταγραφών στη γονιδιωματική ρύθμιση

Abstract

dc:description

The emerging technological developments during the past decade enable large scale analyses in the “regulatory RNA” field and have turned non-coding RNA (ncRNA), initially considered as junk, into a research goldmine. ncRNAs play a crucial role in a remarkable variety of physiological and pathological biological processes. The vast production of data has also been the most important factor underlying the accelerated growth of bioinformatics, a field dedicated to the analysis of data and the development of computational tools indispensable for handling, manipulating and interpreting the results. This thesis focuses on the thorough aggregation of high-throughput data and state-of-the-art Machine Learning techniques in order to develop algorithms for the functional characterization of non-coding transcripts.The current dissertation is specialized on a specific category of RNA transcripts, the microRNAs. microRNAs (miRNAs) are small single stranded non-coding RNA molecules, ~22 nucleotides long, that are loaded into Argonaute (AGO) to induce target cleavage, degradation or translational suppression. Accurate characterization of their targets is considered fundamental to elucidate their regulatory roles. Over the last 15 years, a multitude of in silico and experimental procedures have been developed aiming to determine the miRNA interactome. Currently, high-throughput techniques have enabled the identification of novel experimentally-supported miRNA-gene interactions in a transcriptome-wide scale. This wealth of information is dispersed in a great number of publications and raw datasets. During this thesis DIANA-TarBase v8.0, a reference database devoted to the indexing of experimentally-supported miRNA targets, was designed. Its 8th version is the first database to index more than 1 million entries, corresponding to ~700,000 unique miRNA target pairs, supported by more than 33 experimental methodologies, applied to 592 cell types/tissues under ~430 experimental conditions.AGO-CLIP-Seq experiments are the most widely used high-throughput methodologies. PAR-CLIP variant against AGO proteins methodology has been performed to map miRNA-gene interactions on a transcriptome-wide scale for healthy or disease cell types. Computational methods devoted to AGO-PAR-CLIP present reduced ability to distinguish a large portion of genuine miRNA-targets. To this end, one of the aims of this thesis was to revisit, identify and address current obstacles in AGO-CLIP-Seq analysis. An in silico framework for CLIP-guided identification of miRNA interactions, microCLIP model, was developed. microCLIP is the first relevant implementation to employ the innovative super learner ensemble framework and the only available A-to-Z computational approach for the analysis of AGO-PAR-CLIP datasets. It operates on every AGO-enriched cluster, providing previously neglected functional miRNA binding events with strong RNA accessibility.microCLIP deployment emboldened the development of a next generation de novo miRNA target prediction algorithm. Even the extensive production of relevant approaches observed during the past few years, leading implementations still achieve a far from perfect predictive accuracy followed by an increased number of false positives predictions. Therefore, microT Super Learning framework is presented that maintains and upgrades the pipeline adopted in microCLIP, by enhancing the training with even more high-throughput experiments under a tissue-specific scheme. The new model characterizes interactions with stronger functional efficacy and correctly detects 1.5-fold more experimentally validated target sites when juxtaposed against leading computational approaches. The increased performance of microCLIP and microT frameworks in the detection of miRNA interactions, uncovers previously elusive regulatory events and miRNA-controlled pathways.During this thesis, the candidate participated in 9 scientific studies, involving computational approaches for determining the activity of non-coding transcripts and in two of them is first author. The candidate’s main research activity and contribution in the publications incorporates the implementation of algorithms and automated pipelines for the analysis of Next Generation Sequencing data, data integration for the elucidation of non-coding RNA function and their involvement in mechanisms of post-transcriptional gene regulation. The studies are published in international journals of high impact factor and a total of 942 citations have been received so far, according to Google Scholar.

Degree

thesis:*
Grantor dc:publisher
University of Thessaly (UTH)
Year dc:date
2019

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Karagkouni, Dimitra
  • Καραγκούνη, Δήμητρα

Subjects

dc:subject × 17

Rights

Language dc:language
gre

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/45135

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Karagkouni, Dimitra; Καραγκούνη, Δήμητρα. Υπολογιστική ανάλυση των λειτουργιών των μη κωδικών μεταγραφών στη γονιδιωματική ρύθμιση. University of Thessaly (UTH), 2019. http://hdl.handle.net/10442/hedi/45135