The thesis begins with a detailed introduction to the world of microRNAs (miRNA), focusing on their historic discovery in 1993 and the progress made in understanding their biogenesis mechanisms over the next three decades. The work then transitions to various prediction approaches: it first centres on miRNA themselves, and then on their specific targets. Within this context, prediction methods can be classified into two main categories. The first category encompasses on rule-based methods (ab inito methods) grounded in current knowledge about miRNAs. The second category includes methods that rely on the power of machine learning. This also encompasses the latest approach to prediction, which involves the use of deep neural networks for more precise and effective forecasting. The prediction accuracy varies among individual methods. Nevertheless, for achieving full legitimacy and reliability of predictions, it is of paramount importance that these defined miRNAs are experimentally verified and validated.
|
