Research On Several Key Methods And Application Of Biological Network

Biomedical networks have become one of the main objects in biological studies.In the past century, biological research dominated by reductionism focused onindividual cellular components and their functions, and provided a wealth ofknowledge. Meanwhile, it is increasingly clear that proteins rarely play their rolealone. Instead, most of cellular state changes arise from altered expression of multiplegenes. For example, very often multiple genes collectively contribute to the etiologyand clinical manifestations of most complex diseases. Spurred on by advances inhigh-throughput experimental techniques and the development of omics, researchersare able to observe the expression data of thousands genes or proteins ongenome-wide level. The object of biological researches therefore transformed fromsingle gene to gene set with specific function. However, gene set could not reveal theunderlying mechanism. In order to explore the organization and the cause-and-effectrelationships between genes, people tend to employ molecular interaction networks todescribe cellular status. For example, disease networks are used to represent thepathological mechanism and drug effection network and drug metabolism networksare employed to characterize the cellular response to drug. Biological networksbecome the new―favour‖for biological studies.In this thesis, we attemped to apply methods of complex networks to addressseveral key biological issues, such as construction and functional analysis of complexdisease networks, miRNA-mediated virus-host interactions and crosstalk betweensinaling transduction pathways. Firstly, we introduced biological network inferencemethods based on high-throughput experimental data. Employing knowledge-basednetwork inference algorithm we constructed the distinct network in the brain ofSAMP8mice, and discussed key genes and pathways significantly involved in theaccelerated senescence mechanism of SAMP8mice. Then, we proposed aknowledge-based characterization and functional analysis of biomedical networksnamed as network fingerprint. The relationship between disease networks andpathways was studied based on network fingerprint, and we clustered the diseasesaccording to their network fingerprint. Nextly, we studied mi-RNA mediated virus-host interaction using association network strategy. The connective mapbetween miRNA-virus based on overlap target and we discussed4typical interactionpatterns between miRNA-viral protein host targets. We integrated protein interactioninformation and protein quantification information employing perturbation dynamicsmodel, and explored the relationship between protein function and the abundancetogether with network topology information. Finally, according to the idea ofinterdependence network, we investigated the crosstalk between KEGG signalingtransduction pathways. Especially, the potencial influence of pathway crosstalk ondrug design was discussed. Combining multiple singnaling pathways, we constructedthe global signaling transduction network, and detected possible feedforward loopsand feedback loops completely.Biological networks present genes involved in particular cellular process (node)and regulatory relationship between them (edge) clearly, and they are used to describethe cellular status and reveal the underlying mechanism. In the first chapter, wesummarized the background of biological network studies, and posed several cirticalissues in thes areas. Wen then developed our technical route, and the main contentsand organization of this paper were presented.Biological molecule networks could be measured directly according tohigh-throughput experimental platforms, such as PPI network based on yeasttwo-hybrid technique. However, most networks are inferred from the exprimental dataemploying computing methods, such as gene transcription regulation networks andgene coexpression networks. In the second chapter, we knowledge-based networkinference method was employed on the basis of PPI network, coexpression networkand MetaCore network. We constructed feature sub-networks for hippocampus andcortex of SAMP8mice, and the functional annotation was studied. We identified keynetwork modules and discussed how SAMP8mice can be used as the appropriateanimal model of AD. Our work provides useful clues to understand the underlyingneuro-pathogenesis.It is of chanllenge to interpret biological networks. Given a disease network,researchers can not point out the biological processes mostly associated. In the thirdchapter, we introduced a knowledge-based characterization of biomedical networkscalled―network fingerprint‖. According to measure the similairity between theinferred network and basic network modules (such as pathways), the network wascharacterized by a network fingerprint consisting of the similarity vector. We established the network fingerprint for44disease networks, and studied therelationship between disease and pathways. These disease networks were clusterdbased on their network fingerprints, and the relationship between diseases wasdiscussed.With the rapid development of―omics‖and the accumulation of experimentaldata, association analysis strategy integrating multiple types of data is widely used. Inparticular, it provides new perspective for drug side-effect prediction and drugreposition research. In the fourth chapter, by integrating current data regarding thevirus–human interactome and human miRNA-target pairs, the complex regulatorynetwork between viral proteins and human miRNAs was explored at the system level.We discussed the typical patterns of miRNA mediated virus-host interaction, andfound the overlap between targets of viral proteins and human miRNAs representtopologically important proteins. Our results suggested that miRNAs play importantrole in virus-host interaction, and provided effective clues for further study. Moreover,according to the genes identified in the second chapter, we explored the role ofmiRNAs for SAMP8mice, and discussed the key miRNAs involved in theaccelerated senescence mechanism of SAMP8mice.The interactions between cellular moleculars are dynamical and temporalspecifical. However, the PPI network and gene regulatory networks are static. Thedynamical characteristics of proteins are more important in protein functionexploration and drug target analysis. In the fifth chapter, we employed perturbationdynamics model to integrate the quantitative proteomics of11human cell lines andinteractomes. We found most proteins could lead to very limited perturbation to thenetwork, reflecting the robustness of biological networks. We introduced proteinconcentration potential energy (CPE), and found that CPE was significantly positivecorrelated with protein perturbation factor. The hub proteins in PPI network wasgrouped based on CPE. These results showed the perturbation model could simulatethe real situation of drug action, which will have a wide application on drug targetdesign and drug repositioning.The same as genes rarely play roles alone, biological networks areinterdependence. For example, there are complex crosstalks between signalingtransduction pathways, which cell’s nonlinear response to external stimuli. It willfurther affect multiple biological processes, such as cellular immunity, metabolismand drug design, especially small-molecule kinase inhibitors. A key scientific question is how many crosstalks are there, is the―divide and rule‖strategy feasible forbiological study. In the sixth chapter, we integrated118KEGG signal transductionpathways, and studied the possible effects introduced by pathway crosstalksystematically. We discussed the effects of crosstalks which drug targets involved in.Moreover, we identified two important network motifs: feedback loops andfeedforward loops in the integrated signal transduction network.Finally, a brief summary was introduced, and further possible researches wereprospected.