Studies On Identification Of Potential Therapeutic Targets For Complex Diseases And MiRNA Pharmacogenomics Based On Multi-omics Data And Network Models

Several problems challenge drug discovery and development for complex diseases,e.g.low drug efficiency,severe drug resistance,adverse drug reactions and the demand of personalized medicine.Thus,the single-target-based drug discovery paradigm has been transformed to multiple-target-based paradigm,where innovative methods,new-model tools are in great need to systemically understand the pathology of diseases and assess the drug effects in the whole complex networks.Reccently,with the development of sequencing technologies,multi-omics data are in explosion,especially involving microRNAs,interactome and diseasome,which offer a solid basis for data mining.Biology network model,as an important branch of multi-omics data mining methodology,has developed for almost ten years.Several methods for multi-omics data,such as interactome network building,network topological features analysis,network robustness and control measurement,link prediction,etc.,offer good opportunities to solve the key problems in complex disease area,and to find potential therapeutic targets and stretagy.In this dissertation,we focused 'on two aspects,i.e.prioritizing potential therapeutic targets for complex diseases and studying biomarkers for phamacogenomics,by multi-omics data mining and network medine methodology.Several helpful predictive models and tools were built for complex disease networks.The first chapter generally introduced the backgrounds and overall schedule.The main part of this dissertation,from Chapter 2 to Chapter 5,was mainly divided into two sections.In the first section,two novel methods or tools were developed to find key nodes in the complex networks,which helped to identify potential therapeutic targets for complex diseases.In Chapter 2,an important problem was focused on:the causal relationships between genes and diseases in the complex biological omics network.Firstly,a comprehensive network of protein-protein interactions was built by data integration,which covered almost 85%of the whole protein omics.Secondly,the greedy articulation protein removal(GAPR)method was developed,by applying the algorithm of articulation point identification to the constructed interactome network.GAPR deconstructed the complex disease networks step by step to the residue gaint bicomponent(RGB).Articulation proteins(AP)were identified,which played critical roles in maintaining the network robustness.Through GAPR,the protein-protein interaction network was decomposed into 13 layers and 1 RGB,while the proteome was classified into three types:AP,SP(Supporting Proteins)and RGB proteins.Thirdly,articulation proteins were statistically enriched in the disease-related genes and the drug targets(p<0.05).Last but not least,articulation proteins showed good application ability in the type 2 diabetes mellitus case study to find potential therapeutic targets.GAPR method offered a useful tool to study the key therapeutic targets and their relationships for complex diseases,especially in deep layers.In the third chapter,the miRNA-mediating disease pathology was fucsed on.The measurement of the overall controllability of miRNA on the disease gene regulatory network,helped to find potential disease therapeutic miRNA targets.Firstly,the comprehensive miRNA-mediating gene regulatory network was built by data integration.Network control centrality(Cc)was developed,which measured the achievable maximum match graph in gene regulatory network by controlling one miRNA.Cc helped to identify the prioritized disease-related miRNAs,by systemically assessing the power of each miRNA's controllability on the targeting gene regulatory networks.11 various lung diseases were analyzed,which showed the difference and specificity of the miRNAs' controllability on different lung diseases.Further,differentially expressed miRNAs statistically have higher controllability compared with non-differentially expressed miRNAs in Asthma and Preeclampsia patient case studies.As a measurement of network control,control centrality offered a reference tool to identify potential diagnosis and therapeutic miRNA targets from the aspect of complex disease network control.The second section focused on the interactions between miRNAs and chemicals,which played important roles in disease etiology,development and therapy.In precision medicine era,studying miRNA pharmacogenomics helps to identify potential miRNA biomarkers for diagnosis and therapy of complex diseases.In the fourth chapter,miRNA-mediating environmental toxicology and disease etiology were studied.Predictive environmental factors,miRNAs,diseases association model(PEMDAM)was built as a computational systems toxicology framework.The associations among environmental factors,miRNAs and diseases were systemically analyzed involving miRNA-mediating environmental toxicology and disease etiology.The PEMDAM framework was built using the network-based inference algorithm and incorporating chemical structural similarity and disease phenotypic similarity.Ten computational models built showed high predictive ability in the ten-fold cross validation,and good application ability in case studies of breast cancer and tobacco.The PEMDAM computational systems toxicology framework offered a powerful tool to study miRNA-mediating environmental toxicology and disease etiology.The predicted list can be a reference to design further miRNA experiments.In the fifth chapter,miRNA pharmacogenomics was systemically studied and the effects of drugs were assessed from the whole network aspect to identify potential disease miRNA biomarkers.Firstly,the predictive small molecule,miRNA network-based inference model(SMiR-NBI)was constructed based on a high-quality bipartite network connecting small molecules and miRNAs,which displayed good performance in the ten-fold cross validation.Secondly,the functions of miRNAs were analyzed through their targeting gene networks by network and bioinformatics tools.Thirdly,case studies of natural products and non-steroidal anti-inflammatory drugs identified several potential anticancer mechanisms.Finally,the newly predicted miRNAs for tamoxifen and metformin were experimentally validated in the MCF-7 and MDA-MB-231 breast cancer cell lines via the qRT-PCR assays and yielded high success rates of 60%and 65%,respectively.The online webserver of SMiR-NBI combined all the results and offered a powerful tool for further miRNA pharmacogenomics studies.The sixth chapter made conclusions of the whole dissertation.