Methods For Construction And Analysis Of MicrRNA And Gene Functional Networks

Biological network, as a complex network, is one of the research hotspots inthe fields of information science and systems biology in recent years. Life systemthat constituted by various biological molecules and their interactions can beessentially simplified as a gene network. A gene network is a complex biologicalnetwork that consists of a group of genes (or their products) and the interactionsbetween them. The same group of genes can be moduled in different networks, suchas protein-protein interaction network (PPIN), gene regulatory network (GRN),genetic interaction network (GIN), etc. Based on functional association, functionalnetwork can integerate multiple interaction types to combine various gene networks.Functional interaction not only exists between the genes, but also exists widely inthe miRNAs, which regulate genes’ expression. A large number of biologicalexperiments have promoted the studies on the function of miRNAs and theirinteractions at the system level. Therefore, this dissertation, taking soybean as theresearch object, based on the constructions and analyses of functional gene networkand miRNA network, mainly focuses on the construction and module detection ofmiRNA-gene bilayer network, to dig potential biological knowledge ofmiRNA-gene bilayer functional network, aiming at further revealing the mechanismand characteristics of miRNAs in the control of gene networks at the system level.Firtly, inspired by cell differentiation, dedifferentiation and cell totipotency, theShortest Semantic Differentiation Distance (SSDD) method was proposed tocalculate the semantic similarity between GO terms based on the concepts ofsemantic differentiation and totipotency. Evaluations based on human rating andstandard data shows the excellent performance of SSDD; by further analysis, SSDDis shown not only independent on other datasources, but can also solve the commonproblems of same annotation and shallow annotation, as well as no bias onannotation richness.Secondly, because of the absence of system level functional genome data ofsoybean, the integration of a variety of genome, transcriptome and comparativegenomic data to infer functional gene network can not reproduce in soybean, thoughit is widely used in model organisms. In this dissertation, the functional correlationsof soybean genes were inferred based on GO functional annotations, and thesoybean functional gene networks of four different annotation types (SoyFGN) wereobtained, which cover more than70%of soybean genes deposited in currentdatabases. Topological analysis indicates that SoyFGN shows the typical featuresand modular structures of scale-free biological networks. Based on the KEGGpathway and co-expression data, evaluations indicate that the specially construction of a soybean gene network rather than generating a simple network derived fromother homologous is very important. The network-based resistance gene predictionand analysis indicated that SoyFGN shows effective in gene functional annotation asthe same as that of model organisms, can accurately reflect the gene functionalinteractions at the system level and has good prediction ability.Thirdly, the limited miRNAs are considered to coregulate more genes than theirnumbers through interaction, which opens a new direction for the studies in the fieldof miRNA: studying miRNA interaction networks at the system level by means ofbioinformatics and computational methods. This dissertation presents a method toconstruct the miRNA network based on their functional similarities. In fullyconsideration of the target gene interactive information and regulation strength, thedissertation presents a new miRNA functional similarity measurement, improvingthe measurement accuracy and sensitivity. With the help of functional similaritiesbetween soybean miRNAs and the threshold selection method based on clusteringcoefficient, this dissertation constructed the first soybean sytem-level miRNAfunctional interaction network (SoymiRFN), covering more than90%of the totalsoybean miRNAs been found. Topological analysis indicates that the SoymiRFN isof typical biological network characteristics and modularity. The miRNA functionalnetwork plus functional gene network and miRNA-gene targeting informationconstitute the miRNA-gene bilayer network, which is the first study to use bilayernetwork model to describe the interactions of miRNAs and genes.Fourthly, module detection is an effective network partition method for mininga local structure with more biological significance in a global network. According tothe miRNA-gene bilayer network has features different from the traditionalsingle-layer biological networks or other complex networks, this dissertationpresents a pseudo-3D clustering algorithm, which can realize the identification ofbilayer network modules with hierarchy, overlap and high cohesion, and canautomatically determine the optimal module partition based on the potentialdistribution of current network, while no need to input the number of modules. Thepseudo-3D algorithm was used for the functional module recognition of soybeanmiRNA-gene bilayer network, generating the module partition results withhierarchical structure and overlap. The topological and functional enrichmentanalyses on the optimal modules confirmed the theoretical validity of the pseudo-3Dalgorithm. With the help of experimental data retrieval and literature collection, thein-depth analysis of the soybean fat biosynthesis associated bilayer networkprovides experimental evidence for the effectiveness of the pseudo-3D algorithm,and reveals the characteristics of soybean gene and miRNA related to lipidbiosynthesis.Lastly, based on the above four research contents and results, to make up for the insufficiency of current functional soybean genome databases, this dissertationestablished a database specifically for soybean functional network-SoyFN. Thedatabase provides a rich and friendly interface for retrieval, visualization, analysisand download soybean gene-gene, miRNA-miRNA and miRNA-genes functionalinteraction information and other omics information. In addition, SoyFN databaseintegrates the soybean functional omics data provided from KEGG, UniprotKB,SoyBase, EnsemblPlants and other well-known public databases. At the same time,it also provides very practical analysis tools, such as SoySearch, Genome Browserand ID mapping. Using examples show that the SoyFN can provide a completeseries of analysis processes of soybean functional gene network, and is useful forresearches related to both soybean experimental science and bioinformatics area.The database can be accessed at http://nclab.hit.edu.cn/SoyFN/by any browsersupporting HTTP protocol.