| Microorganisms widely exist in various environments,such as ocean,soil and human body,and are closely related to human production and life.The development of high-throughput sequencing’s technology helps us to quickly and accurately obtain the relevant information of the microbiome,which promotes the research in the field of microbiology.However,the uniqueness and complexity of microbial data,especially the high sparsity of data,bring challenges to statistical analysis and interpretation.The problem of zero value processing in microbial data has always been an important research content in Biostatistics.Microbial data is a special kind of composition data.This paper tries to apply the zero-value processing method in composition data to zero-value processing of microbial data to test its application effect.First of all,the zero-value replacement methods of some component data have certain requirements for the proportion of zero values in the data,and the sparsity of microbial data makes these methods unusable.We selected zero-value replacement methods for component data that can be used to process microbial data,and three commonly used zero-value processing methods in biology,and applied them to the simulated data of case group and control group,and compared the selected methods in error detection rate and efficacy by using difference analysis.The results show that in most cases,the zero-value replacement method of component data is better than the common zero-value replacement method of biological group data.In addition,we also compare the method of not replacing zero value with all the previous methods.We find that the method of not replacing zero value is better than the zero value replacement method of component data and the common zero value replacement method of microbial data.This is somewhat different from our expectation,but it is reasonable.This may be because the zero value can be regarded as a special observation value,and replacing the zero value destroys the original structure of the data itself to some extent,so the effect of not replacing the zero value is the best.Finally,we apply the findings of this paper to real microbial data.Without replacing the zero value,the species related to Ulcerative Colitis and Crohn’s disease were identified by difference analysis,and compared with the related species obtained by other scholars.The results were not only highly consistent with the existing research results,but also found some new species that might be related to the disease.Our findings have enriched the species related to diseases to some extent,which may bring new research directions for the prevention and treatment of diseases. |
