Local machine learning-based feature importances for gene regulatory network inference

Abstract

Understanding how a cell (or organism) reacts to a change in the environment or disturbance requires an understanding of the intricate processes controlling gene expression and, therefore, protein synthesis. A common representation of these mechanisms is the gene regulatory network, that aims at defining the regulation links between genes as a set of interactions. Inferring those gene regulatory networks from expression data has been a widely studied field at the level of bulk expression data. However, recent breakthroughs in sequencing technologies enables measurements at the resolution of a single cell. Such data allows the development of research towards the analysis of gene regulatory networks for a single specific cell or for a distinct cell type, rather than global interactions. This thesis has the objective to perform these analyses.