Circadian rhythms, i.e. rhythms exhibiting a cyclic behaviour with a period of approximately 24 hours, are present in the metabolism of most living organisms. The transcriptional processes, i.e. the processes associated with mRNA synthesis, critically contribute to their origination, and are responsible for most of the mechanisms which regulate gene expression levels in cells. Inhibition or activation of a putative transcriptionally regulated ‘child’ gene can be achieved via binding of proteins called transcription factors (TFs) to the gene promoter, a region of the DNA containing protein-specific binding sites. In this work, we investigate modelling and inference approaches for different scenarios of circadian transcriptional regulation. We focus on a system which comprises two transcription factors and a regulated child gene. We first perform parameter inference in the context of state-space models on simulated data from a mechanistic stochastic model describing this scenario. Additionally, we investigate the effect of data aggregation across different cells, and derive the smoothing equations for a destructive sampling scenario. In the second part of this work, we consider a situation in which an important regulator of a child gene has not been observed. We apply our model to mRNA expression levels of a subset of circadian genes of the Arabidopsis Thaliana model plant. Inference is in this case aimed at estimating both the model parameters and the unobserved transcription factor profile. We compare a posteriori the inferred transcription factor profiles with available time-series data for one important circadian regulator in the Arabidopsis Thaliana, namely late elongated hypocotyl (LHY), and identify similarities for a several genes known to belong to the central clock. Finally, we focus on a scenario of transcriptional regulation which includes an auto-regulatory negative feedback loop. This modelling framework is motivated by the availability of spatio-temporal imaging data of genes belonging to the mammalian central clock in mice suprachiasmatic nucleus (SCN), and in particular here we focus on Cry1. We introduce a distributed delay to account for nuclear export, translation, protein complex formation, and nuclear import, of the molecular species involved. To perform inference, we develop a novel filtering algorithm that can be applied to any system with distributed delays. We finally apply the methodology to Cry-luc spatio-temporal data, and find that parameter estimates are spatially distributed, with a marked difference between central and peripheral SCN regions.
| Identifer | oai:union.ndltd.org:bl.uk/oai:ethos.bl.uk:714921 |
| Date | January 2016 |
| Creators | Calderazzo, Silvia |
| Publisher | University of Warwick |
| Source Sets | Ethos UK |
| Detected Language | English |
| Type | Electronic Thesis or Dissertation |
| Source | http://wrap.warwick.ac.uk/88571/ |
Page generated in 0.0022 seconds