Study On The Structure Learning Technology Of Microbial Network Based On Network Embedding

The analysis of microbial networks can help to discover the interactions of microbial communities,which are crucial to understanding their building mechanisms,structural functions,and identifying the core species within them.However,traditional microbial network structure learning techniques rely too much on individual associations between nodes,and it is difficult to capture higher-order relationships that composed more global features.In order to solve the above problems,this paper introduces the network embedding technology and proposes a structure learning method of microbial network based on network embedding technology.The main research contents of this paper include the following points:(4)Analysis of module stabilit: By calculating the Gini coefficient of embedded module,traditional first-order module and single strain,it is proved that embedded module is more stable than single strain and traditional first-order module.(5)Regression prediction analysi: The regression prediction model based on XGBoost is constructed,and the embedded module,the traditional first-order module and the abundance of single strain in the environment are used as the characteristic input to predict the indexes of various environmental factors,proving that the introduction of embedded module is helpful to improve the performance of the prediction model.(3)Analysis of module composition difference: The difference in composition between the embedded module and the traditional first-order module is compared by calculating the Jaccard similarity.It is proved that the microbial community characteristic information contained in the embedded module is different from the traditional first-order module,and the two complement each other.(2)Correlation analysis of environmental factor: The correlation between embedded module,traditional first-order module,single strain and environmental factors is studied,and it is proved that the embedded module and the traditional first-order module has a stronger correlation with environmental factors than OTU.(1)Build high-order network module: The microbial similarity network is constructed based on the real data set,and the network is mapped to the low-dimensional space using the network embedding method to obtain the low-dimensional vector representation of each microorganism.By spectral clustering of these vectors,the embedded module,namely the high-order network module,is finally obtained.The results show that the structure learning technology of microbial network based on network embedding can effectively extract the higher-order relationships of microbial networks.And high-order network modules created by network embedding can be served as a potential new biomarker for feature extraction of microbial communities.