Computational Integrative Models for Cellular Conversion: Application to Cellular Reprogramming and Disease Modeling

The groundbreaking identification of only four transcription factors that are able to induce pluripotency in any somatic cell upon perturbation stimulated the discovery of copious amounts of instructive factors triggering different cellular conversions. Such conversions are highly significant to regenerative medicine with its ultimate goal of replacing or regenerating damaged and lost cells. Precise directed conversion of damaged cells into healthy cells offers the tantalizing prospect of promoting regeneration in situ. In the advent of high-throughput sequencing technologies, the distinct transcriptional and accessible chromatin landscapes of several cell types have been characterized. This characterization provided clear evidences for the existence of cell type specific gene regulatory networks determined by their distinct epigenetic landscapes that control cellular phenotypes. Further, these networks are known to dynamically change during the ectopic expression of genes initiating cellular conversions and stabilize again to represent the desired phenotype. Over the years, several computational approaches have been developed to leverage the large amounts of high-throughput datasets for a systematic prediction of instructive factors that can potentially induce desired cellular conversions. To date, the most promising approaches rely on the reconstruction of gene regulatory networks for a panel of well-studied cell types relying predominantly on transcriptional data alone. Though useful, these methods are not designed for newly identified cell types as their frameworks are restricted only to the panel of cell types originally incorporated. More importantly, these approaches rely majorly on gene expression data and cannot account for the cell type specific regulations modulated by the interplay of the transcriptional and epigenetic landscape. In this thesis, a computational method for reconstructing cell type specific gene regulatory networks is proposed that aims at addressing the aforementioned limitations of current approaches. This method integrates transcriptomics, chromatin accessibility assays and available prior knowledge about gene regulatory interactions for predicting instructive factors that can potentially induce desired cellular conversions. Its application to the prioritization of drugs for reverting pathologic phenotypes and the identification of instructive factors for inducing the cellular conversion of adipocytes into osteoblasts underlines the potential to assist in the discovery of novel therapeutic interventions.