Systems-level control of microRNA regulation through transcriptomic plasticity
Loading...
Date issued
Authors
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Reuse License
Abstract
MicroRNAs (miRNAs) are post-transcriptional regulators of gene expression and are deeply embedded within gene regulatory networks that coordinate cellular identity. By targeting large sets of transcripts and interacting with other regulatory layers, miRNAs contribute to adaptable control of gene expression programs. While high-throughput sequencing has enabled comprehensive profiling of miRNAs across biological contexts, the complexity and context dependence of miRNA-mediated regulation pose major challenges for computational analysis.
This dissertation addresses core computational challenges in miRNA research, ranging from miRNA annotation and target identification to the inference and interpretation of miRNA regulatory activity from transcriptomic and genomic data. A central premise of this work is that miRNA function must be analyzed within the regulatory and structural context of the transcriptome, as transcriptomic plasticity driven by alternative polyadenylation (APA) reshapes miRNA–target interactions and complicates conventional analytical approaches.
To address the high false-positive rates of miRNA target prediction, predicted miRNA–mRNA interactions are integrated with transcription factor regulatory networks. This network-based framework reveals sequence composition and positional binding-site features that distinguish functional targets and highlights the importance of regulatory context in miRNA targeting. Extending this network-based perspective to transcriptomic plasticity, single-cell analyses of glioblastoma show that cell-type-specific APA dynamically reshapes miRNA binding site availability and rewires miRNA–transcription factor co-regulatory networks in association with cellular states in the tumor microenvironment. To overcome the limited availability of small RNA sequencing data, interpretable machine learning models are applied to infer miRNA expression from matched mRNA profiles, enabling reconstruction of miRNA regulatory activity from widely available transcriptomic datasets. Consideration of transcript dynamics further reveals that apparent differential miRNA expression between conditions can arise from target-site dynamics rather than true changes in miRNA abundance, as miRNA decay mechanisms and APA-driven gain or loss of binding sites confound standard differential expression analyses. Complementing these transcriptomic findings, large-scale analyses of human genomic variation link sequence conservation and population-level constraint to miRNA functional relevance and annotation reliability.
Together, this work addresses key challenges in miRNA annotation, expression prediction, identification of targeting features, decay dynamics, and regulation of miRNA activity through alternative polyadenylation. By integrating regulatory network analysis, machine learning, modeling of transcriptomic dynamics, and population genetics, this dissertation provides an integrated, context-aware computational framework for studying miRNA function across complex biological systems.