Optimization And Applications Of Microbial Subspecies Identification Method

Subspecies identification is one of the most critical issue in microbiome studies,as it is directly related with the functions of the species as well as the whole microbial communities in response to the environmental stress and their feedbacks.However,identification of subspecies remains a challenge largely due to the small variances between different strains under the same species.Subspecies identifications' accuracies heavily depend on variant identification and categorization from microbiome data,though current SNP calling from microbiome data remain suspectable under statistical scrutinization.In our research,we developed a new microbial subspecies identification method Strain-GeMS,and verified its performance on simulated data and ab initio data,and also made a preliminary application on in vivo data.The main tasks of the thesis include:(1)We have first made a survey on the existing microbial community classification methods,especially the methods for subspecies/species identification.By comparing the modules of these methods,we understand their advantages and disadvantages.(2)Then based on better SNP calling tools and better clustering methods,we developed a new subspecies identification method Strain-GeMS.In the method,we first used Metaphlan2 to identify the species in which subspecies analysis could be performed,then used optimized MultiGeMS to call SNPs,followed by hierarchical clustering of these SNPs and generating all subspecies models with SNP-flow,finally used PAM clustering to determine the optimal subspecies combination model.Results on simulated,and ab initio datasets have shown that Strain-GeMS could always outperform other subspecies identification methods.(3)We preliminarily applied Strain-GeMS on large scale of in vivo metagenomic data and discovered the subspecies difference at different time points or different environment.With the rapidly increasing amount of microbiome samples,and the needs for subspecies identification,we believe that Strain-GeMS could become a key tool towards elucidating of subtle differences among subspecies in a microbial community.