Research On Identification Method Of Biological Network Module Based On Labels

With the rapid development of bioinformation technology,more and more bioomics data such as gene,protein and metabolism have been mined and applied.Networks can well describe the structure of these complex biological systems.The module characteristic of the network is very important for people to analyze the potential data connection of the system.Module recognition is a method used to find a set of nodes in a network that are closely connected internally but sparsely connected externally.However,general module recognition methods only focus on the topology of the network and ignore the labels of nodes and the weights of edges,which makes network information not fully utilized and affects the accuracy of module recognition results to some extent.In addition,the traditional method has some limitations in detecting overlapping modules,one node is only allocated to one module.In many real networks,modules do not exist independently,but overlap to some extent.Therefore,the module recognition method considering overlapping structure has more practical research value.Aiming at the above problems,this paper proposes a weighted multi-label propagation algorithm WNMLPA(Weighted Node-similarity-based Multi-Label Propagation Algorithm)based on CESNA Algorithm and NMLPA Algorithm.The WNMLPA method obtained edge weights by modeling and optimizing the network topology and node attribute information,and applied the weighted sum of edge weights and node attribute similarity to the multi-label propagation algorithm for module identification.The WNMLPA method not only integrates node information and edge information,but also allows the detection of overlapping modules.In the experiment,LFR reference network,real network and biological network data sets were selected,and F1-score,Jaccard Similarity Coefficient and modularity were taken as measurement indexes to compare WNMLPA method with current relatively advanced Attributed SBM,NMLPA and CESNA methods.Experimental results show that compared with other methods,the module recognition results obtained by WNMLPA method are closer to the real division of the network,and the module structure is also clearer.This paper also presents a visual system to show the biological network visually.Users can choose a variety of layout ways to observe the internal structure characteristics of the network,or through data screening to separate out the subnet of interest for further research.In addition,the system also supports the query of node biological function,which provides convenience for people to analyze the network.