A Data-driven Discovery System for Studying Extracellular MicroRNA Regulation and Trafficking

MicroRNA is a class of short non-coding RNAs (~22bp long) that play critical roles in post-transcriptional gene regulation. In humans, microRNAs endogenously regulate more than 60% of human genes and participate in most biological and pathophysiological processes. Meanwhile, microRNAs secreted into circulation are considered as important cell-to-cell communication mediators. While many studies focus on characterizing the role of extracellular microRNA in human disease both experimentally and computationally, such functional analysis hinges on addressing challenging questions about how and which microRNAs can be transferred and how they interact with functional genes. Particularly, computational solutions to microRNA target prediction have been hindered by the fact that the microRNA-gene interaction is conditionally specific where competitive binding is involved. This dissertation mainly introduces three new methodologies for studying the regulation and trafficking of extracellular microRNA, along with a suite of software tools. Our first contribution improves microRNA target prediction by integrating competitive microRNA regulation through a meta-LASSO model. Particularly we investigate the dynamic and conditional properties of the microRNA-gene regulation network and its association with tumor progression through statistical modeling along with the integration of genomics, transcriptomics, and epigenetics data from multiple human cancers. In this study, we innovatively utilized sequencing data to quantitatively estimate the regulatory potential of a microRNA to all possible targets, which significantly reduced the false positive rate. Secondly, a classification model was built for identifying human absorbable exogenous microRNAs using a Manifold Ranking approach, as well as a large set of discriminative features selected using an SVM-based recursive feature eliminator. Lastly, a novel graph-based motif finding method entitled MDS2 was developed to effectively detect sequence motifs among extracellular microRNAs that are potentially associated with the microRNA sorting and trafficking process. Unlike other motif finding tools, MDS2 integrates a large-scale sampling process to assess k-mer enrichment based on stringent statistical significance. These methodologies are either one of a kind, or outperform other state-of-the-art methods according to experimental evaluations. Various data analytics, modeling, and data mining techniques will also be discussed in this dissertation