Constructing Temporal Transcriptional Regulatory Cascades in the Context of Development and Cell Differentiation

Abstract

Cell differentiation is a complex process orchestrated by sets of regulators appearing at precise temporal points, resulting in regulatory cascades that affect the expression of broader sets of genes, ending up in the formation of different tissues and organ parts. The identification of stage-specific master regulators and the mechanism by which they activate each other is a key to understanding and controlling differentiation and still a challenging quest, particularly in the fields of tissue regeneration and organoid engineering. To tackle this quest I developed a novel workflow and a model I call the Temporal Regulatory Cascade (TRC). The TRC workflow combines a comprehensive general regulatory network based on binding site predictions with user-provided temporal gene expression data, to generate a series of connected stage-specific regulatory networks. The TRC identifies those regulators that are unique for each time point and the regulatory interactions between them, taking into consideration the temporal order of their appearance. The TRC model is represented in the form of a regulatory cascade that shows the emergence of these regulators and regulatory interactions across time. The TRC workflow was implemented in the form of a user-friendly tool with a visual web interface that requires no expert knowledge in programming or statistics, making it directly usable for scientists with no strong computational background. In addition to generating TRCs, the tool links multiple interactive visual workflows, in which a user can track and investigate further different regulators, target genes, and interactions, directing the tool along the way into biologically sensible results based on the given dataset. The workflow was used to analyze a high-quality dataset that documents the gene expression levels across multiple time points during the differentiation of stem cells into mature cardiomyocytes. In addition to the main dataset, we applied the TRC model to several different time-series expression datasets coming from different contexts such as neural development. The model was successful in identifying previously-known and new potential key regulators, in addition to the particular time points to which these regulators are associated. These results were highly supported by GO enrichment, experimental knowledge and literature. Compared to other methods, our approach showed an advantage in terms of computational time, and the density of the important regulators identified in such small cascades. The workflow is now available publicly at TF-Investigator.sybig.de/TRC.