Network Inference from Multi-omics and Phenotypic Profiles

基於多組學和表型數據的網絡推斷

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date13 Jan 2021

Abstract

In recent years, ‘Big data’ has been used to describe the rapid increase in volume, variety and velocity of information in almost every aspect of our lives, the biomedical research is also not an exception. With the continuous emergence of novel sequencing and genome editing technologies, the massive data of multi-omics (e.g. genome, transcriptome, epigenome, proteome and metabolome) and of genetic modification is generating worldwide. The complexity of these biological data poses a need of systematic methodology to analyze and interpret. Therefore, computational systems biology, which aims to develop and apply efficient algorithms for the goal of quantitative modeling and interpreting of biological systems, is becoming critical in biological studies. Actually, this interdisciplinary field keeps making significant progress to excavate the biological data to address fundamental questions in biology, which is expected to further assist the clinical decision making and drug discovery.

For computational systems biology, network inference is one primary methodology to analyze large quantities of biological data by modeling the interactions among the biological molecules. Gene regulatory network (GRN) refers to the network formed by the interaction between genes in a cell (or within a specific genome). In cancer study, the heterogeneity is a challenge to precisely discover pathogenesis and conduct treatments. Focusing on the cancer subtypes with a poorer prognosis, based on regulatory network inference using ‘RTN’ and further computational dissection of latent regulatory mechanisms, the master regulatory elements as well as their target genes could be identified as potential prognostic and predictive biomarkers. Another important problem related to GRN and cancer subtype specificity is dissecting cancer subtype-specific signaling pathways, which is crucial to pinpointing dysregulated genes for the prioritization of novel therapeutic targets. Nested effects models (NEMs) are specifically tailored to reconstruct signaling networks from indirect observations of genetic perturbations. We propose to extend the original NEMs to predict drug targets by incorporating causal information of (epi)genetic aberrations for signaling pathway inference. Except for the directed gene regulatory or signaling network, another type of biological network is the association network inferred by the correlation of expression or functional signatures among genes or macromolecules. In order to systematically dissect the functional association from image-based phenotypic screening profiles of bacterial mutants, we design to detect functional modules from the inferred posterior association network (PAN). Besides the PAN inference and module detection strategy within the computational framework, an image processing flow for medium-scale assays of bacterial mutants based on the open-source R language is developed.

This thesis focuses on the research about the computational system biology methods for excavating regulatory mechanisms of cancers and searching for functional modules of microorganisms. The content in each chapter is described below:

Chapter 1: An introduction to computational system biology is given. Basic concepts and properties of the network, the typical types of biological networks, and representative reconstruction methods of gene regulatory networks are briefly reviewed. In addition, the study objectives are summarized.

Chapter 2: This chapter describes a specific experimental and computational framework to detect functional modules from the posterior association network inferred by image-based phenotypic screening profiles of virulence-associated Pseudomonas aeruginosa mutants. Two functional modules related to QS and T6SS processes were detected and further exploration about the genetic interaction was conducted.

Chapter 3: This chapter introduces the identification of subtype-specific master regulatory microRNAs in pancreatic ductal adenocarcinoma based on regulatory network inference and integrative computational dissection of latent regulatory mechanisms. We identified two master regulatory miRNAs, miR-29c and miR-192, regulating TGFβ signaling pathway and LOXL2, ADAM12 and SERPINH1, target genes of miR-29c, showing strong association with prognosis.

Chapter 4: In this chapter, we have developed NEM-Tar, which extends the original nested effects models (NEMs) to predict drug targets by incorporating causal information of (epi)genetic aberrations for the signaling pathway inference. Also, an information theory-based score, weighted information gain (WIG), was proposed to assess the impact of signaling genes on downstream reporter genes. Simulation studies and two case studies were performed for NEM-Tar.

Chapter 5: This chapter talks about summarization and perspective.