Use this URL to cite or link to this record in EThOS:
Title: Resolving cell fate : experimental, computational and mathematical methods in single cell transcriptomic analysis
Author: Jawaid, Wajid
ISNI:       0000 0004 8501 0135
Awarding Body: University of Cambridge
Current Institution: University of Cambridge
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Deeper understanding of the embryological origins of tissues and organs is likely to provide insights into novel clinically relevant preventative and therapeutic strategies. To do so effectively and at a large scale so as to have clinical significance requires an exhaustive and meticulously accurate knowledge of normal morphological development and the underlying molecular pathways. A variety of lineage tracing and classical molecular biology techniques have led to key insights but emerging technologies now offer the possibility of more fine grained and precise measurements at the level of the single cell rather than ensembles of heterogeneous cells. In this study the rapidly developing technology of single cell RNA sequencing was combined with the development of state of the art computational methods to study murine gastrulation and early organogenesis at a hitherto unprecedented granularity. A comprehensive analysis of FLK1+ cells harvested from gastrulating murine embryos was performed. In-silico reconstruction of pseudo-temporal and pseudo-spacial relations were demonstrated. Using prior knowledge of known driver genes deeper substructure was revealed in clusters that were initially defined by unsupervised algorithms, illustrating that careful implementation of supervised approaches can outperform naïve unsupervised methods. In this way multiple members of the leukotriene branch of the arachidonic acid pathway were found to be enriched within a subset of the endothelial cluster that has a molecular signature consistent with yolk sac derived definitive wave haematopoiesis. This was subsequently validated in an in-vitro embryonic stem cell differentiation colony assay. By developing a novel adaptation to tSNE dimensionality reduction that now allows new data points to be mapped backed to previously calculated embeddings, the FLK1+ data set became a reference to which cells from other experiments were mapped. The Tal1-/- knockout mutant was characterised, reaffirming the known phenotype with complete failure of embryonic haematopoiesis. Analysing the Tal1-/- endothelial cluster shows definitively in the in-vivo organism no activation of an alternative cardiac fate programme as was previously postulated from an in-vitro model system. Additionally tSNE mapped Brachyury cells into the void in the FLK1+ dataset and identified a node like population. Loss of spacial context remains an Achilles' heel of single cell protocols. Generation of the single cell suspensions leads to disruption of cellular contacts and loss of any spacial information. A novel method is described which uses tSNE of bulk spacial data to pre- condition a tSNE map upon which single cells can then be computationally positioned reconstructing spacial context. To model gene interactions and perturbations from single-cell data, a hybrid feed-forward deep neural network was trained on branching pseudo-temporally arranged single cell qPCR data of in-vivo wild-type murine developmental haematopoiesis. Strikingly despite the model never having 'seen' a mutant, in-silico gene perturbations in the deep neural network are able to faithfully reproduce the Tal1-/- phenotype. In summary the use of single cell transcriptomics to probe early murine embryology com- bined with development of new methods has uncovered a novel pathway in embryonic haematopoietic development and allowed in-silico reconstruction of a short period of early embryonic haematopoiesis. Critically these methods have broad application within the fields of developmental and stem cell biology.
Supervisor: Gottgens, Berthold ; Nichols, Jennifer Sponsor: Wellcome Trust
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
Keywords: Single cell ; Transcriptomics ; 10X Genomics ; sequencing ; high-throughput ; neural network ; tSNE ; diffusion maps ; gaussian process ; developmental biology ; embryology ; haematopoiesis ; stem cell ; cell fate ; fate ; differentiation ; waddington ; landscape ; PCA ; leukotriene ; LTC4 ; Flk1 ; blood ; endothelium ; heart ; cardiac ; transcription termination variants ; 3 prime ; 3' ; sequence capture ; drop out