Uncovering the system-level regulatory architecture of gene expression in humans

Abstract

Uncovering the system-level transcriptional regulatory architecture of gene expression in human has been a major focal point in modern systems biology researches. Human genes are commonly under coordinated and combinatorial transcriptional regulation mediated by a class of proteins known as transcription factors. Recent technological advancements have enabled comprehensive mapping of transcription factor binding motifs in human genome through both experimental (e.g. ChIP-seq) and computational (e.g. comparative sequence analysis) methods. Collections of defined transcription factor binding motifs and their corresponding target genes in a given species can be used as a "dictionary" to define groups of genes that have the potential of being under coordinated control. Utilising a large-scale transcription factor binding motif location data in human genome, this study has defined a comprehensive genome-wide binding motif dictionary of human transcription factors. Instead of inferring transcriptional regulatory networks from co-expression, this study attempted to take the opposite approach in which, by using simple multivariate data analysis methods and graph theory, groups of genes under putative transcriptional co-regulation were defined based on the promoter motif content similarity between genes. These defined networks of genes could be used as a platform for incorporating other biological datasets to clarify the system-level regulatory architecture of human genes. In this thesis, independent gene function annotation, gene expression and ChIP-seq datasets were employed to bring in functional and biological insights, and also to provide independent validations for the regulatory networks defined based on promoter motif content similarity. The results showed the regulatory networks defined in this work were of biological significance. More specifically, groups of genes under putative co-regulation defined in this study on the basis of sharing common transcription factor binding motifs were likely to share common function and expression. Overall, the affirmative findings described in this thesis demonstrated the feasibility of identifying putative gene regulatory networks by using large-scale motif dictionaries. Show more