Date of Award
Doctor of Philosophy (PhD)
Mathematics, Statistics and Computer Science
MicroRNAs (miRNAs) are small noncoding transcripts that can regulate gene expression, thereby controlling diverse biological processes. Aberrant disruptions of miRNA expression and their interactions with other biological agents (e.g., coding and noncoding transcripts) have been associated with several types of cancer. The goal of this dissertation is to use multidimensional genomic data to model two different gene regulation mechanisms by miRNAs in cancer. This dissertation results from two research projects. The first project investigates a miRNA-mediated gene regulation mechanism called competing endogenous RNA (ceRNA) interactions, which suggests that some transcripts can indirectly regulate one another's activity through their interactions with a common set of miRNAs. Identification of context-specific ceRNA interactions is a challenging task. To address that, we proposed a computational method called Cancerin to identify genome-wide cancer-associated ceRNA interactions. Cancerin incorporates DNA methylation (DM), copy number alteration (CNA), and gene and miRNA expression datasets to construct cancer-specific ceRNA networks. Cancerin was applied to three cancer datasets from the Cancer Genome Atlas (TCGA) project. We found that the RNAs involved in ceRNA interactions were enriched with cancer-related genes and have high prognostic power. Moreover, the ceRNA modules in the inferred ceRNA networks were involved in cancer-associated biological processes. The second project investigates what biological functions are regulated by both miRNAs and transcription factors (TFs). While it has been known that miRNAs and TFs can coregulate common target genes having similar biological functions, it is challenging to associate specific biological functions to specific miRNAs and TFs. In this project, we proposed a computational method called CanMod to identify gene regulatory modules. Each module consists of miRNAs, TFs and their coregulated target genes. CanMod was applied on the breast cancer dataset from TCGA. Many hub regulators (i.e., miRNAs and TFs) found in the inferred modules were known cancer genes, and CanMod was able to find experimentally validated regulator-target interactions. In addition, the modules were associated with distinguishable and cancer-related biological processes. Given the biological findings obtained from Cancerin and CanMod, we believe that the two computational methods are valuable tools to explore novel miRNA involvement in cancer.