Research On The Methods And Applications For Integrated Analysis Of Multiple Biomedical Networks

Since the molecular equipment in human cells generally possess functional interaction relationships, our understanding of disease etiology has gradually changed from a single gene mutation to the intracellular and intercellular network disturbance of body tissues and organs. The emerging network medicine not only provide us the platform to explore complex molecular mechanism of a particular disease, which helps to characterize the molecules and pathways associated with diseases, also let us have the chance to find the relationships between molecules which were implicit in apparently different(pathological) phenotypes. The integration analysis of multiple networks will help us to discover the organizational nature of biological networks, the analysis of network structure and system dynamics between multiple networks will reveal the inner relationship of these networks, thereby resulting a more comprehensive understanding to the whole life system. Under the guidance of mathematical theory, a multiple biomedical networks analysis framework is constructed, and the numerical characteristic of biomedical network is also build. This paper studies the problem of multi-network alignment and network fusion, two applications of related integration methods are given.With the rapid development of international scientific projects such as LINCS, TCGA and ENCODE, the myriad disease-relevant data of various omics are sustained growing. With the booming of network science methodology and constantly updating bioinformatics methods, a huge amount of approaches accumulated in biological researches provide the basis for either a comprehensive analysis of relationships in different omics or the discovery of important biological laws. Based on the emerging network science methods, we carried out the research on the integration analysis of multiple biomedical networks.Biological network researches put forward a series symbolic concept for biology, including the node degrees, small-world, scale-free, etc. These concepts well describe the properties of biological networks only from one of its aspects, but the systematic language of mathematics which could completely characterize biological networks is still lack. Firstly, our research applied the ideas of numerical functional theory to build a mathematical expression of biological network. Based on inner product space, we proposed the concept of biological network space to operate the network objects. Then, under the background of biological network space, the projection algorithm was applied to the numerical characterization of biological networks, which laid a solid theoretical foundation for the establishment of network characterization standards.Biomedical researchers are usually difficult to understand the definitions of complex molecular networks, as they are unfamiliar with the complicated and abstract math notions. To decipher the molecular network in a way that will be familiar to most biomedical researchers, we proposed a characterization method of biomedical networks based on biological knowledge by comparing the target network to the classic “basic networks”. In our method, the target network was projected into a “spectrum-like” long vector by calculating the similarity between target network and basic networks, called network fingerprint. The knowledge-based multidimensional characterization provides a more intuitive way to parse the molecular network, especially for large-scale network comparison and cluster analysis. As an example, we studied and analyzed the network fingerprint of 73 disease network extracted from the KEGG database. By comparing the fingerprints of these disease networks, we explored the connections among diseases and signaling pathways. Our analysis suggested that the network fingerprint method is an effective method to decipher biological network.Organism is an organic whole composed by a variety of molecules, as the biological network constructed by interacting between molecules is of great importance to the discovery of operation of biological system. The multi-network analysis model, which could stimulate the complicated live systems, has significant implications on a thorough understanding of principle of molecular biology. On the basis of network fingerprint method we provide a web application called Network Fingerprint Scanner(NFPScanner) to characterize and compare biological network based on classic biological knowledge. By representing biological network as its network fingerprint, NFPScanner provides the ability to do network functional analysis in three levels: network fingerprint, network alignment and network enrichment. Biologists could find the potential laws of biological systems by adjusting different reference network datasets during the network fingerprint analysis in NFPScanner. By finding optimal subnet mapping, NFPScanner could help explore the function correlation among network modules. NFPScanner could also visualize GO enrichment and pathways enrichment.Recent omics-technologies have made it effective to obtain diverse types of genome-wide data. To depict a panorama of a given disease or a biological process, computational methods which could integrate these data are imperative. Network fusion method is a good way to absorb the data tsunami generated by various omics, and play important role in the conversion of complex multidimensional data into refined biological knowledge, even laws. Network fusion method is greatly superior, in respect of efficiency, to those analyses with single data type. Meanwhile network fusion method provides an integrated analysis approach on the identification of disease subtype, has high clinical value.With the long-term development of multiple omics analysis, multidimensional biomedical data also continues to present a very intensive growth. Therefore, the integrated clustering analysis of various omics data is very important to develop individual medicine and precision medicine. However, while method for clustering analysis has made rapid progress, the developed methods are lacking an intuitive web interface that assists the biomedical researchers without enough programming skill. To fill the gap in practical applications, we proposed a web tool named Integrated Clustering of Multi-dimensional biomedical data(ICM), which provides an interface from which to fuse, cluster, and visualize multidimensional biomedical data and knowledge. In ICM, users could explore the heterogeneity of a disease or a biological process by identifying the subtype of patients. The results of ICM can be interactive revised on the intuitive and user-friendly interface. Through the exchange of the network links corresponds to a given project ID, researchers could share the results analyzed in ICM with their collaborators, thereby directly promote results sharing. ICM also supports the incremental clustering analysis, which means the user could directly add new samples in the existing analysis project to obtain the new clustering results.Finally, a brief summary was introduced, and further possible researches were prospected.