Genes indispensable for proper embryonic development show intricate patterns of expression
throughout the time, space and magnitude of their activity. This diversity is enabled by
elaborate regulatory mechanisms that guide their expression. They also possess a distinct
type of core promoters that enable the integration of all regulatory inputs. However, it is still
not clear how is coordination of regulation achieved. The first step towards understanding
this process is to characterise dynamics of expression, and core promoter features that process
the regulation.
In this thesis, I explored the diversity of spatio-temporal gene expression during zebrafish
development. I defined a novel measure of anatomical specificity that defines how precisely
an anatomical structure is defined in the Anatomical Ontology system. Using anatomical
specificity measure, I quantified gene expression dynamics from mRNA in situ hybridisation
data. Gene expression divergence from in situs was used to predict expression levels from
RNA-seq expression data. This analysis allowed me to propose a measure of gene expression
complexity which showed that genes with the highest complexity score are developmental
genes, whereas genes with low complexity score are involved in housekeeping functions. Next,
I developed a method that reports significantly enriched core promoter elements in a group
of genes. Using this method, I compared differences in core promoter composition in active
genes expressed in different developmental periods. In addition, this method found groups
of genes with a specific core promoter structure that are specified for a biological process.
Finally, I used scRNA-seq data from zebrafish development to identify patterns of gene
co-expression across different cell clusters. Co-expression suggests that a gene pair possesses
a common regulatory programme. I show that genes with the most divergent co-expression
patterns across development are developmental genes and that housekeeping genes have least
diverse co-expression patterns. I went further to create co-expression networks which allowed
me to analyse co-expression patterns into more details.