As cancer tumours develop, competition between cells will favour those with some mutations over others, creating a dynamic heterogeneous system made up of different cell populations, called sub-clones. This heterogeneity poses a challenge for treatment, as this variety serves to ensure there is almost always a portion of the cells which are resistant to any one targeted therapy. This can be avoided by combining therapies, but finding viable combinations experimentally is expensive and time-consuming. However, there is also cooperation between sub-clones, and being able to better model and predict these dynamics could allow this interdependence to be exploited. In order to investigate how best to tackle tumour heterogeneity, while avoiding acquired resistance, I have developed the first comprehensive computational model of the gene regulatory network in breast cancer focused on the c-myc oncogene and the differences between sub-clones. I model the system as a discrete, qualitative network, which can reproduce the conditions in heterogeneous tumours, as well as predict the effect of perturbations mimicking mutations or application of therapy. Together with experimental collaborators, I apply my computational model to an in vivo mouse model of MMTV-Wnt1 driven breast cancer, which has high and low c-myc expressing sub-clones. I show that the computational model is able to reproduce the behaviour of this system, and predict how best to target either one sub-clone individually or the tumour as a whole. I show how combination therapies offer more paths to attack the tumour, and how two drugs can work synergistically. For example, I predict how Mek inhibition will preferentially affect one sub-clone, but the addition of COX2 inhibition improves effectiveness across the tumour as a whole. In this thesis, I show how a computational network model can predict treatments in breast cancer, and assess the effects on different clones of different treatment combinations. This model can be easily extended with new data, as well as adapted to different types of cancer. This therefore represents a novel method to find viable combination therapies computationally and speed up the development of new cancer treatments.

A hallmark of the joint disease osteoarthritis (OA) is the degradation of the articular cartilage in the affected joint, debilitating pain and decreased mobility. At present there are no disease modifying drugs for treatment of osteoarthritis. This represents a significant, unmet medical need as there is a large and increasing prevalence of OA. Using a systems biology approach, we aimed to better understand the pathogenic mechanisms of OA and ultimately aid development of therapeutics. This thesis focuses on the analysis of gene expression data from human OA cartilage obtained at total knee replacement (TKR). This transcriptomics approach gives a genome-wide overview of changes, but can be challenging to interpret. Network-based algorithms provide a framework for the fusion of knowledge so allowing effective interpretation. The PhenomeExpress algorithm was developed as part of this thesis to aid the interpretation of gene expression data. PhenomeExpress uses known disease gene associations to identify relevant dysregulated pathways in the data. PhenomeExpress was further developed into an 'app' for Cytoscape, the widely used network analysis and visualisation platform. To investigate the processes that occur during the degradation of cartilage we examined the gene expression of damaged and intact OA cartilage using RNA-Seq and identified key altered pathways with PhenomeExpress. A regulatory network driven by four transcription factors accounts for a significant proportion of the observed differential expression of damage-associated genes in the PhenomeExpress identified pathways. We further explored the role of the cytokines IL-1 and TNF that have been reported to β drive the progression of OA. Comparison of the expression response of in vitro cytokine-treated explants with the in vivo damage response revealed major differences, providing little evidence for any significant role of IL-1 and TNF as drivers of OA β damage in vivo. Finally, we examined the heterogeneity of OA through analysis of cartilage expression profiles at TKR. Through a network-based clustering method, we found two subgroups of patients on the basis of their gene expression profiles. These subgroups were found to have distinct OA expression perturbations and we identified TGF and S100A8/9 β signalling as potentially explaining the observed differential expression. We developeda RT-qPCR based classifier that allowed classification of new samples into these subgroups so allowing future assessment of the clinical significance of these subgroups. The work presented in this thesis includes a novel, widely-accessible tool for the analysis of disease gene expression data, which we used to give new insights into the pathogenesis of osteoarthritis. We have produced a rich dataset for future research and our analysis of this data has increased our understanding of cartilage damage processes and the heterogeneity of OA.

Conventional drug discovery focuses on single protein targets and follows a “sequence, structure, and function” paradigm for selecting best protein targets to screen lead chemical compounds. This established paradigm simply avoids addressing directly the challenge of evaluating chemical toxicity and side effects until a later stage of drug discovery, resulting in inefficiencies and increased time and cost. We developed a new “network biology” perspective to assess proteins as potential drug targets using emerging biomolecular network data sets. To do so, we integrated several types of biological data for current drug targets from DrugBank, protein interaction data from the HAPPI and HPRD databases, literature co-citation data from PubMed, and side effects data from FDA-approved drug usage warnings. We used the Bayes factor and Positive Predictive Values to examine the use of certain network properties, such as network node degrees and essentiality, to predict candidate drug targets. We also developed a metric to evaluate a protein target’s overall side effects by taking into account aggregated side effect scores of all FDA-approved drugs targeting the protein. We discovered that non-essential protein with lower-to-medium network node degree could better serve as drug targets when combined with conventional protein function information. Integrated biomolecular associations, instead of physical interactions, are better sources for predicting drug targets with network biology methods. Our network biology framework presents exciting promises in developing better drug targets that lower the side-effects at later stages of drug development and help establish the field of “network pharmacology.”

This work focuses on the use of network graph theory in biological networks. I explore how network graph theory informs our understanding of biological networks such as protein interaction networks. I show that the human protein interaction network forms dynamic, modular structures that organize cell signaling pathways into higher order units. The misregulation of the dynamic, modular structure of the protein interaction network in breast cancer tumours is associated with outcome of the disease, suggesting that the altered structure of the protein interaction network is directly related to the phenotype of the tumour. I also demonstrate that the human protein interaction network is fractal in nature and thus forms self-similar structures within the network. The fractal skeletons of the protein interaction network contain critical information and therefore can be used alone in determining the phenotype of breast cancer tumours by examing the disruption of dynamic network structures. The self-similar fractal backbones deconvolve the protein interaction network into layers of independent function, resulting in improved description of breast cancer outcome using the dynamic network modularity algorithm. Finally, I discuss how the discoveries and technologies described within can be improved and how these discoveries can lead to a network based modality of medicine.

This work focuses on the use of network graph theory in biological networks. I explore how network graph theory informs our understanding of biological networks such as protein interaction networks. I show that the human protein interaction network forms dynamic, modular structures that organize cell signaling pathways into higher order units. The misregulation of the dynamic, modular structure of the protein interaction network in breast cancer tumours is associated with outcome of the disease, suggesting that the altered structure of the protein interaction network is directly related to the phenotype of the tumour. I also demonstrate that the human protein interaction network is fractal in nature and thus forms self-similar structures within the network. The fractal skeletons of the protein interaction network contain critical information and therefore can be used alone in determining the phenotype of breast cancer tumours by examing the disruption of dynamic network structures. The self-similar fractal backbones deconvolve the protein interaction network into layers of independent function, resulting in improved description of breast cancer outcome using the dynamic network modularity algorithm. Finally, I discuss how the discoveries and technologies described within can be improved and how these discoveries can lead to a network based modality of medicine.

Research into human disease has classically been 'bottom-up', focussing on individual genes. However, the emergence of Systems Biology has prompted a more holistic 'top-down' approach to decoding life. Less than a decade since the complete draft of the human genome was published, we are increasingly in a position to model the interacting constituents of a cell and thus understand molecular perturbations. Given biological systems are rarely attributable to individual molecules and linear pathways, we must understand the complex dynamic interplay as cellular components interact, combine, overlap and conflict. The integrative approach afforded by Network Biology provides us with a powerful toolset to understand the vast volumes of omics data. In this thesis, I investigate both infectious disease, specifically HIV infection and heritable disease. HIV, the causative agent of AIDS, represents an extensive perturbation of the host system and results in hijacking of cellular proteins to replicate. I first introduce the HIV-interaction data and then characterise HIV's hijack, revealing the ways Network Biology can greatly enhance our understanding of host-pathogen systems and ultimately the systems itself. I find a significantly greater propensity for HIV to interact with ''key'' host proteins that are highly connected and represent critical cellular functions. Unexpectedly, however, I find there are no associations between HIV interaction and inferred essentiality and genetic disease-association. I hypothesise that these observations could be the result of ancestral selection pressure on retroviruses to minimise interactions with phenotypically crucial proteins. Investigating inherited disease, I apply a similar integrative approach to determine the relationships between inherited disease, evolution and function. I find that 'disease' genes are not a homogenous group, and that their emergence has been ongoing throughout the evolution of life; contradicting previous studies. Finally, I consider the consequence of bias in literature-curated interaction datasets. I develop a novel method to identify and correct for ascertainment bias and demonstrate that failure to do this weakens conclusions. correct for ascertainment bias and demonstrate that failure to do this weakens conclusions. The aim of this thesis has been to explore the ways Network Biology can provide an integrative biological approach to studying infectious and inherited disease. Given billions of people around the world are susceptible to disease, it is ultimately hoped that a Systems Biology approach to understanding disease will herald new pharmaceutical interventions.

Traditional medicines (TM) have been used for centuries to treat illnesses, but in many cases their modes-of-action (MOAs) remain unclear. Given the increasing data of chemical ingredients of traditional medicines and the availability of large-scale bioactivity data linking chemical structures to activities against protein targets, we are now in a position to propose computational hypotheses for the MOAs using in silico target prediction. The MOAs were established from supporting literature. The in silico target prediction, which is based on the “Molecular Similarity Principle”, was modelled via two models: a Naïve Bayes Classifier and a Random Forest Classifier. Chapter 2 discovered the relationship of 46 traditional Chinese medicine (TCM) therapeutic action subclasses by mapping them into a dendrogram using the predicted targets. Overall, the most frequent top three enriched targets/pathways were immune-related targets such as tyrosine-protein phosphatase non-receptor type 2 (PTPN2) and digestive system such as mineral absorption. Two major protein families, G-protein coupled receptor (GPCR), and protein kinase family contributed to the diversity of the bioactivity space, while digestive system was consistently annotated pathway motif. Chapter 3 compared the chemical and bioactivity space of 97 anti-cancer plants’ compounds of TCM, Ayurveda and Malay traditional medicine. The comparison of the chemical space revealed that benzene, anthraquinone, flavone, sterol, pentacyclic triterpene and cyclohexene were the most frequent scaffolds in those TM. The annotation of the bioactivity space with target classes showed that kinase class was the most significant target class for all groups. From a phylogenetic tree of the anti-cancer plants, only eight pairs of plants were phylogenetically related at either genus, family or order level. Chapter 4 evaluated synergy score of pairwise compound combination of Shexiang Baoxin Pill (SBP), a TCM formulation for myocardial infarction. The score was measured from the topological properties, pathway dissimilarity and mean distance of all the predicted targets of a combination on a representative network of the disease. The method found four synergistic combinations, ginsenoside Rb3 and cholic acid, ginsenoside Rb2 and ginsenoside Rb3, ginsenoside Rb3 and 11-hydroxyprogesterone and ginsenoside Rb2 and ginsenoside Rd agreed with the experimental results. The modulation of androgen receptor, epidermal growth factor and caspases were proposed for the synergistic actions. Altogether, in silico target prediction was able to discover the bioactivity space of different TMs and elucidate the MOA of multiple formulations and two major health concerns: cancer and myocardial infarction. Hence, understanding the MOA of the traditional medicine could be beneficial in providing testable hypotheses to guide towards finding new molecular entities.

Neurodegeneration is a chronic, progressive and debilitating condition that affects majority of the World's elderly population who are at greater risk. Numerous scientific studies suggest that there could be a common underlying molecular mechanism that promotes the degeneration and the subsequent neuronal loss, however so far the progress in this direction is rather limited. Abnormal protein misfoldings, as well as protein plaque formations in the brain, are some of the hallmark characteristic features of neurodegenerative disorders (NDDs). Genetic and environmental factors, oxidative stress, excessive reactive oxygen species formation, mitochondrial dysfunction, energy depletion and autophagy disruption etc. are some of the widely suspected mechanisms that manifest the cognitive, motor and emotional symptoms of these NDDs. Motivated by some molecular traits found in common in several NDDs, network-based systems biology tools and techniques were used in this study to identify critical molecular players and underlying biological processes that are common for Parkinson's, Alzheimer's and Huntington's disease. Utilizing multiple microarray gene expression datasets, several biomolecular networks such as direct interaction, shortest path, and microRNA regulatory networks were constructed and analyzed for each of the disease conditions. The network-based analysis revealed 26 genes of potential interest in Parkinson's, 16 in Alzheimer's and 30 in Huntington's disease. Many new microRNA-target regulatory interactions were identified. For each disorder, several routes for possible disease initiation and protection scenarios were uncovered. A unified neurodegeneration mechanism network was constructed by utilizing the significantly differentially expressed genes found in common in Parkinson's, Alzheimer's and Huntington's microarray datasets. In this integrated network many key molecular partakers and several biological processes that were significantly affected in all three NDDs were uncovered. The integrated network also revealed complex dual-level interactions that occur between disease contributing and protecting entities. Possibilities of microRNA-target interactions were explored and many such pairs of potential interest in NDDs were suggested. Investigating the integrated network mechanism, we have identified several routes for disease initiating, as well as alleviating ones that could be utilized in common for Parkinson's, Alzheimer's and Huntington's disease. Finding such crucial and universal molecular players in addition to maintaining a delicate balance between neurodegeneration promoters and protectors is vital for restoring the homeostasis in the three NDDs.

L'interférence par l'ARN est un processus selon lequel un petit ARN non codant se lie à un ARN messager cible dans la cellule pour moduler son expression. Ce mécanisme a été conservé au cours de l'évolution : il est retrouvé aussi bien chez les animaux que chez les végétaux. Nous savons aujourd'hui que le rôle de l'interférence par l'ARN est fondamental, dans le développement embryonnaire comme dans la progression tumorale. Les microARN (miARN) sont des ARN non codant endogènes dont l'une des particularités est leur capacité à réguler tout un ensemble de gènes par interférence avec les ARN messagers. Il est ainsi prédit qu'un seul miARN serait capable de réguler plusieurs centaines de gènes différents. La thèse a consisté en l'analyse de la corégulation médiée par les miARN grâce à l'inférence de réseau basée sur le partage de gènes cibles. La corégulation est un phénomène où plusieurs miARN différents interviennent sur les mêmes familles de gènes et donc sur les mêmes processus biologiques. Le travail a plus spécifiquement consisté en la mise en place d'un réseau de miARN, en son analyse topologique mais également en son interprétation biologique. Le but final était de proposer de nouvelles hypothèses biologiques à tester afin de mieux comprendre la corégulation des processus biologiques par les miARN. Au travers de ces travaux, deux groupes de miARN ont pu être mis en évidence, dont l'un impliqué dans la régulation de la signalisation par les petites GTPases – hypothèse par la suite validée par plusieurs expériences in vitro. Dans un second temps, une communauté de miARN impliquée dans le maintien de la pluripotence des cellules souches a pu également être mise en évidence. Pour compléter ces analyses, une étude systémique de la topologie des réseaux de miARN a été menée afin de mieux comprendre leur intégration dans les réseaux biologiques et leur rôle dans le devenir cellulaire. / RNA interference is a process in which a small non-coding RNA will bind to a specific messenger RNA and regulate its expression. This evolutionary conserved mechanism is found in all superior eukaryotes from plants to mammals. Nowadays, we know that RNA interference is a major regulatory process involved in developmental biology and tumor progression. MicroRNAs (miRNAs) are endogenous (coded in and produced by the cell) non-coding RNAs which are able to regulate a whole set of genes, typically hundreds of genes. This doctoral thesis consisted in the analysis of the miRNA mediated coregulation through a network approach based on target sharing. Coregulation is the process where many different miRNAs will regulate the same set of genes and thus the same biological process. In particular, the work consisted in the inference of a miRNA network, in its topological analysis and also its biological interpretation. Indeed, the final aim of the work was to generate new biological hypothesis. As such, two different groups of miRNAs were first retrieved. One of them was predicted to be involved in the small GTPase signaling and was further validated in vitro. Moreover, a miRNA community involved in the maintenance of stem cells pluripotency was also discovered. Finally, a systemic analysis of the target-based miRNAs network was conducted to better understand their integration with biologic networks and their role in cell fate.

Cancer is the second most common cause of death in the United States. It is a complex disease with environmental, genetic, and lifestyle factors influencing the likelihood of getting cancer and the development of any resulting tumor. Understanding the genetics of cancer is integral to developing novel patient-specific treatments. However, due to complexity, hundreds to thousands of tumors are required for sufficient power to identify the network of relationships among these genes. Animal models of cancer are commonly used to reduce cost and to control experimental variables allowing for more specific hypothesis testing. The Sleeping Beauty transposon mutagenesis system can be used to model cancer in mice. While the Sleeping Beauty mutagenesis system is an important tool in understanding cancer, it has specific computational needs. Experiments need to be analyzed in a fast, unbiased, and efficient manner. A computational method must also accurately model the system allowing for validation and interpretation. Here I present an updated Integration Analysis System and use this system to validate the assumptions present in forward genetic screens of cancer using the Sleeping Beauty. This system allows for rapid identification of cancer genes, but does not directly aid in understanding the relationship between the genes. Given the complexity of cancer, understanding the relationship between cancer genes is very difficult. I have created a connectedness network utilizing the STRING database to better derive an understanding of cancer genes. STRING is a database of known and predicted protein-protein interactions. The connectedness between pairs of genes is calculated using a network reliability metric. This database allows for increased power to detect known pathways when compared to STRING alone. Combining this connectivity network with the set of cancer genes identified by the Integration Analysis System is a strategy for rapid and efficient interpretation of the genetic results.