Computational Approaches to Investigate Alternative Polyadenylation in Single-Cell RNA-Sequencing Data

ABSTRACT

Alternative polyadenylation (APA) is a critical mechanism in post-transcriptional regulation, influencing gene expression by generating mRNA isoforms with distinct 3′ untranslated region (3′ -UTR). Dysregulation of APA has been associated with various diseases, including neurodegenerative disorders such as Amyotrophic Lateral Sclerosis (ALS). In this thesis, I investigate APA dysregulation in ALS by leveraging single-cell RNA sequencing (scRNA-seq) technologies to profile cell-type-specific APA dynamics in ALS. I developed a hybrid convolutional neural network with multi-head attention (CNNMATT) model to explore APA landscapes in different brain cell types across both sporadic and familial ALS cases, revealing novel insights into APA profiles that may contribute to ALS pathology. Additionally, I investigate the broader role of RNA-binding proteins (RBPs) in APA regulation using whole-genome Perturb-seq data, identifying previously uncharacterized regulators of APA. This research culminates in the identification of novel APA regulators, broadening our understanding of APA’s regulatory framework in disease. Overall, this work delivers a cell-type-specific view of APA dysregulation in ALS and a genome-wide survey of APA regulators via Perturb-seq, jointly laying the groundwork for future mechanistic and therapeutic investigations into RNA processing across diverse biological contexts.