Abstract
Molecular interaction, such as Protein-Protein Interaction (PPI), Protein-DNA Interaction (PDI), and Genetic Interaction (GI), plays important roles in development and maintenance of organisms. This thesis addresses the problems of assessing PPI confidence, inferring molecular interaction network for cell differentiation, and analyzing how the molecular interaction network controls cell cycle.In Caenorhabditis Elegans (C. elegans), a large number of PPI has been identified by either small-scale (e.g. co-immunoprecipitations) or high-throughput (e.g. yeast two-hybrid) experiments. However, the confidence of the PPIs detected from these different types of experiments varies. Therefore, it is necessary to weigh the confidence of each PPI from the experiments. Currently, a comprehensive weighted PPI network is not yet available in C. elegans. In the first part of this thesis, an integrative PPI network in C. elegans is constructed with 12951 interactions involving 5039 proteins from seven molecular interaction databases. A reliability score based on probabilistic graphical model (RSPGM) is proposed to assess PPIs. The main parameter of RSPGM score contains a few latent variables which can be considered as several common properties between two proteins. Validations on four high confidence yeast PPI databases show that RSPGM provides more accurate evaluation than other approaches. Furthermore, this network is employed on inferring interaction path of the canonical Wnt/β -catenin pathway in C. elegans. Six out of eight genes on the inferred interaction path are validated to be Wnt pathway components. Therefore, RSPGM is essential and effective for evaluating PPIs and inferring interaction path. Finally, a user interactive website is built for querying and visualizing the PPI network with RSPGM scores.
Various types of cells compose diverse tissues and organs in the organism. How are these different cell types generated? What is the origin of cell type diversity? In the second part of this thesis, these problems are studied by inferring gene regulatory network for cell differentiation at cellular level. By combining weighted PPI data from the first part of this thesis, PDI data from public, and state-of-the-art single-cell resolution gene expression data from experiments, a framework of regulatory pathway inference is proposed based on probabilistic graphical model. Specifically, gene regulatory mechanisms are studied for two marker genes, pha-4 and nhr-25, which are highly related to pharynx and skin formation respectively. Firstly, gene expression of the two marker genes are quantified at cellular level in C. elegans embryo. Then, the gene regulatory effects are investigated, by comparing wild-type and mutant (RNAi gene knockdown) gene expression data via two proposed statistical hypothesis tests pipelines. In addition, a probabilistic graphical model of pathway inference is constructed. Optimal configurations of model attributes are inferred based on belief propagation method. Finally, directed and signed gene regulatory networks are obtained, which reveal gene regulatory mechanisms for pharynx and skin cell differentiation in C. elegans embryo.
Cell differentiation aims to generate different cell types for specific function, while cell cycle is a very critical process to control cell division which increases cell number. Similar to other biological functions and processes, cell cycle is regulated by a large number of genes. In the third part of this thesis, a cell-cycle network is constructed based on the knowledge of key genes and their interactions from literature studies, to understand how these genes control cell cycle proceeding in C. elegans early embryo. Then, a discrete dynamical Boolean model is applied in computer simulations to study dynamical properties of the network. With the Boolean model, it shows the cell-cycle network is robust, compared with random networks and tested under several perturbations. Moreover, a biological pathway is observed in the simulation, which corresponds to a whole cell-cycle progression. Finally, to investigate whether the proposed network could explain biological experiment results, the network simulation results are compared with gene knockdown experiment data. The smaller number of attractors and the shorter biological pathway from gene knockdown network simulation interpret the shorter cell-cycle lengths in the mutant data from RNAi gene knockdown experiment. This indicates that the proposed network reveals the gene regulatory mechanism of cell cycle in C. elegans early embryo.
To sum up, this thesis contains three parts, assessing PPI to obtain a weighted PPI network for the first time in C. elegans, inferring gene regulatory network to understand the origin of tissue formation in C. elegans embryo, and dynamically analyzing gene regulatory network to reveal how genes control cell cycle proceeding in C. elegans early embryo. Studies of C. elegans systems physiology enable a unique opportunity to infer the gene regulatory network of pharynx and skin cell differentiation and develop the weights PPI network in C. elegans to address common problems such as putative therapeutic targets in cancer treatments.
| Date of Award | 25 Jan 2016 |
|---|---|
| Original language | English |
| Awarding Institution |
|
| Supervisor | L H Leanne CHAN (Supervisor) & Hong YAN (Supervisor) |