Novel computational approaches to research longitudinal microRNA-mRNA expression datasets

Abstract

microRNAs (miRNAs) regulate many biological processes and are used as biomarkers for the classification of diseases, conditions and developmental stages. miRNAs function by targeting and negatively regulating specific mRNAs. One limitation of utilising miRNAs in experimental work is the complex and often redundant behaviour of miRNA-mRNA interactions; as a single miRNA can regulate many mRNAs and one mRNA can be regulated by multiple miRNAs. This complexity stifles the potential of miRNAs. However, miRNAmRNA expression datasets are becoming generated more frequently and they can help to garner greater understanding of how miRNAs regulate biological systems. Furthermore, researchers are generating longitudinal datasets as these can elude to greater understanding of how biological conditions change over time. Thus there is a rise of longitudinal miRNA-mRNA expression datasets. However, extracting useful information from increasingly sophisticated datasets is a challenge in biological research. Exploration of such datasets using computational techniques, such as big data bioinformatics, kinetic modelling and machine learning could help in identifying interesting miRNA-mRNA interactions. During this PhD I asked if these methodologies can be used to gain insights from a range of longitudinal miRNA-mRNA expression datasets. Hence, I developed an R/Bioconductor tool called TimiRGeN to integrate, analyse and generate small networks from longitudinal miRNA-mRNA datasets. Datasets from kidney fibrosis, chondrogenesis dataset, breast cancer and Huntington’s disease (HD) were analysed with TimiRGeN. Results from the chondrogenesis dataset analysis were taken forward to generate a multimiRNA kinetic model. With help from my collaborators this model was validated and predictions were made. Using the HD dataset, machine learning (ML) techniques trained models to detect if samples have disease or wild type conditions. Overall, I have developed and used multiple computational techniques to increase knowledge gained from longitudinal miRNA-mRNA datasets, and I believe the results show these techniques can contribute to miRNA research.