Machine learning methods for MicroRNA target prediction

Phelan, Ryan James (2023) Machine learning methods for MicroRNA target prediction. Doctoral thesis, University of East Anglia.

Preview

PDF Download (8MB) | Preview

Abstract

MicroRNAs are small non-coding RNA molecules that form a post-transcriptional layer of gene regulation. microRNA binds with messenger RNA in order to repress translation and accelerate its degradation, ultimately downregulating the expression of genes. The mechanics of these bindings in animals are complex and entrenched in a myriad of contextual factors which influence the specificity and efficacy of potential interactions.

This thesis describes the development of miRsight, a novel target prediction tool utilising advanced machine learning techniques. miRsight is trained using 44 target recognition features compiled through testing on published microRNA-transfected RNA sequencing data, an experimental procedure in which microRNA molecules are introduced into a sample to quantify their impact on gene expression. In addition to the tool itself, a database of pre-computed predictions is hosted at https://mirsight.info, which also provides search, filter, and export functionality for user convenience.

The results of this study indicate that miRsight is able to more effectively predict and rank microRNA targets compared to popular target prediction tools. This is validated by examining the downregulation of gene expression from predicted targets using microRNA transfection. In the 12 samples reserved for testing, miRsight is shown to more consistently identify true targets in the top 100, 300 and 500 of predictions by rank compared to TargetScan, MirTarget and DIANA-microT. Additionally, miRsight is capable of producing several thousand total predictions for each microRNA while maintaining this high rate of prediction accuracy.

Item Type:	Thesis (Doctoral)
Faculty \ School:	Faculty of Science > School of Biological Sciences
Depositing User:	Nicola Veasy
Date Deposited:	11 Jul 2024 08:25
Last Modified:	11 Jul 2024 08:25
URI:	https://ueaeprints.uea.ac.uk/id/eprint/95859
DOI:

Downloads

Actions (login required)

View Item