Investigation of the gene regulatory logic encoded in genome sequences using machine learning

Spatiotemporal regulation of gene expression in multicellular organisms is mediated by cell-type specific regulatory elements called transcriptional enhancers. Their activity is regulated by the coordinated binding of transcription factors (TFs) to specific DNA sequences encoded within these elements. In this PhD thesis, we investigate transcriptional enhancers in the context of pluripotency and early neural development. In Chapter 2, we investigate how enhancer activity is encoded in the DNA sequence underlying TF binding sites. To this end, enhancer activities of NANOG binding sites in naive mouse embryonic stem cells (mESCs) were measured using ChIP STARR-seq. This revealed two distinct classes of NANOG binding sites, active enhancers and inactive binding sites. Predictive modelling of these classes uncovered sequence features which account for these differences such as the ESRRB motif. In Chapter 3, we examined differential enhancer activities of NANOG binding sites between naive and primed mESC using ChIP STARR-seq. Comprehensive analysis of various chromatin genomic features revealed chromatin accessibility changes as a major predictor of differential enhancer activity. Predictive modelling using only DNA sequence features uncovered potential transcriptional regulators that define enhancer activity in the two cellular states, such as TCF3, TCF4 and ESRRB for naive mESC and ZIC3 for primed cells. In Chapter 4, we studied the timing of neural induction during early nervous system development along the anterior-posterior axis using information about enhancer accessibilities. Changes in open-chromatin states during neural development were measured by applying ATAC-seq to an in vitro model system. Analysis of resulting data revealed that for posterior spinal cord induction, cells initially commit to a posterior regional identity before acquiring neural identity. Moreover, our analysis highlighted the role of CDX2 during this process which represses hindbrain and activates spinal cord related transcriptional programs. Together, this Chapter provides conclusively evidence for the dual origin of neural progenitor cells.Open Acces