Year: 2018
Type: dissertation
Abstract: Instructions from the genome are first copied to make messenger RNAs, which are then translated to make proteins. To expand the repertoire of these instructions, cells can modify the messenger RNAs in different ways. Two such modifications are alternative polyadenylation and alternative splicing. Using RNA-Seq data and deep learning, we trained computational models that can be applied to sequences in the genome to predict tissue-specific polyadenylation and splicing patterns. Presented with multiple alternative polyadenylation sites, the polyadenylation model can predict the probability each site would be selected for cleavage and polyadenylation. Similarly, given alternative exons, the splicing model can predict which exon would more likely be included. The performance of these models in predicting polyadenylation and splicing patterns for genomic regions not observed during training is evaluated, and an analysis of what the models have learned reveals sequence elements that are known to influence these cellular processes. Importantly, these computational models are trained on genome-wide patterns based on the reference genome but can generalize to individual variations. Each model can thus be viewed as a simulator, where the genotype of an individual can be fed in as an input, and the output describes how the individual's mutations affect the mechanisms of polyadenylation and splicing in different tissue types. The relevance of these models for problems in genomic medicine is described, and proof-of-concept applications are demonstrated.
Source: TSpace (University of Toronto)
Author: Michael K. K. Leung
Language: English
Topic: RNA Research and Splicing
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W3150197612