Study On Analysis Method Of Compositional Data In Microbiome By Integrating Phylogenetic Tree Information

Microbial communities play an important role in human health and disease,and the distribution of microbes is usually measured by the relative abundance of species(compositional data),general statistical methods could not apply to such compositional data.To solve the problem of compositional data,scientists proposed three transformation methods:ALR,CLR,and ILR transforms.PhILR introduced phylogenetic data based on the ILR transformation,provided a framework for the analysis of phylogenetic data combined with microbial abundance data and it improved the efficiency of microbiological data analysis.Phylogenetic data recorded the evolutionary relationship of species.PhILR introduced the information of the phylogenetic tree so it can analyze and explain the microbiome from the perspectives of genetic and evolution.This article is mainly based on PhILR framework and has made the following two tasks and innovations:Firstly,using the topology structure of phylogenetic tree to measure the weights of PhILR transformation.PhILR didn't make use of the phylogenetic tree information when weighting the "Balance" after transformaion.This article weights the PhILR transformed weights based on the phylogenetic tree height and the number of sub-nodes of the phylogenetic tree.Nodes with higher height of phylogenetic tree and more child nodes gives greater weight.This idea was used to improve the PhILR method,and the validity of our improvement is demonstrated in the comparison of accuracy in four machine learning classifications with three microbial datasets.Secondly,constructing the microbial networks based on phylogenetic tree information.The PhILR method has already combined the information of the evolutionary tree and the abundance data of the species.However,the new variable after PhILR transformation is no longer a microbe,but a hybrid of many microbes.Using the PhILR transformation to construct networks directly has a problem in the explanation of the network.Based on the transformation relationship between ILR and CLR,this paper transformed PhILR into the form of CLR,and used CLR to calculate the correlation and build the microbial network.In this way,the evolutionary tree information is combined with the abundance data to construct the microbial network,and we can analyse interactions between microbes from the perspective of evolution.It is all along the difficulty in integrating microbial compositional data and phylogenetic tree information.Based on the PhILR framework,this paper integrated the microbial component data and phylogenetic tree information.We improved the PhILR by using the height of the evolutionary tree and the number of sub-nodes,and constructed the microbial network based on phylogenetic information.The experimental results show that the two methods proposed in this paper are effective and feasible.