| Microorganisms are important components of the Earth’s ecosystem.It is crucial for research on human health to study the structure and function of microbial communities,food safety and the environment.With the improvement of sequencing technology throughput,microbiome-related research developed rapidly.However,the cost of using high-throughput sequencing technologies to study the microbiome remains relatively high.Microbiome data are usually high-dimensional and sparse,so there is a possibility of compressibility.If a combination of microbial abundance measurement can be made to obtain composite measurements from which all microbial abundances can be reconstructed,the cost of microbiome sequencing experiments might be significantly reduced.In this paper,we propose a microbiome data collection method based on compressed sensing theory and composite measurement strategy,which can measure microbial abundance compositely and then reconstruct the whole complete microbial abundances from the composite measurement data.The main work of this paper is as follows:First,a microbiome data collection method based on compressed sensing theory and composite measurement strategy is proposed.The simulation experiments of microbial abundance composite measurement and data reconstruction are designed based on multiple microbial abundance datasets.First of all,it designs the microbial abundance composite measurement process;then,try to search the potential module structure in microbial abundance data by microbial module decomposition methods such as probability matrix decomposition and non-negative K-SVD;finally,the abundance of all microorganisms is reconstructed from the lower dimensional composite measurement data using compressed sensing theory.Whether the low-dimensional composite measurements data can represent the high-dimensional original microbial abundance data and the effect of data reconstruction were experimentally investigated.The experimental results show that the proposed microbiome data collection method can effectively reconstruct the original highdimensional microbial abundance data from the low-dimensional microbial abundance composite measurement data.Second,a reconstruction strategy of microbial composite measurement data is proposed,which integrates graph regularized nonnegative matrix decomposition and compressed sensing.Based on the assumption that "if two samples are close in their original space,they should also be close to each other in the mapping space",we can consider the association between different samples to more accurately obtain the dictionary matrix of microbial modules,thus obtaining better data reconstruction results.Specifically,by using graph regularized non-negative matrix decomposition,the geometric relationships between the original data are retained in the decomposed low-dimensional matrix,and the obtained microbial modules are applied to the process of reconstruction.The experimental results show that the proposed integration graph regularized nonnegative matrix decomposition and compressed sensing microbial composite measurement data reconstruction strategy effectively improves the reconstruction of microbial abundance data from low-dimensional composite measurements.This research explores the strategy of composite measurement of microorganisms based on the compressed sensing theory,which provides new ideas for the design of sequencing experiments and other data collection of microbiome,and provides a feasible direction to the improvement of microbial transcriptome imaging and other techniques,and has potential theoretical value for microbiome research. |
