Characterisation of VapC in Mycobacterium smegmatis

Abstract

Members of the VapC family of proteins cleave RNA at specific sites in order to regulate biological processes with a cell. Characterization of the sites targeted by a specific protein using conventional biochemical techniques is resource intensive. This study explores the potential use of computational models to characterize the sites targeted by VapC in Mycobacterium smegmatis. Previous work has reported the impact of VapC upon each gene in the M. smegmatis genome and produced a hypothesis model for the specific motif targeted by the enzyme. However, this model has been shown to be insufficient for the differentiation of sites cleaved by VapC from those not cleaved. This study aims to extend this model to accurately describe the features which influence VapC activity at a site. A model capable of accurately predicting the VapC target sites could supplement the existing biochemical techniques. Furthermore, a process developed to train such a model could potentially be generalized and applied to other proteins and species. This thesis explores increasingly complex representations of RNA sites and a suite of supervised learning techniques to train models that predict the efficiency with which sites are cleaved by VapC. The simplest representations of RNA sites consider only the RNA sequence. More detail is added to the representation in the form of secondary structures and the potential influences of tertiary structures are discussed. No model is produced that is capable of accurate, meaningful predictions. This suggests that the construction of a successful model requires significant alterations to the representation of RNA sites or that the data available is insufficient for training an accurate model.