Abstract:
|
Due to the high cost and limited resources of microbiome experiments, prioritizing the differentially abundant features as candidates for downstream analysis is of high importance. Ranking the differential abundance of microbiome features in a logical order may inform on the full complexity of a biological mechanism rather than using an arbitrary statistical threshold like P-value, and it's difficult to confirm the identified OTUs when a large quantitative of OTUs are declared differentially abundant by an arbitrary statistical threshold. Although some approaches have been developed for ranking genes in RNA-seq data, there is still no method for ranking OTUs in the amplicon-based microbiome data. We propose a hierarchical Zero Inflated Poisson model that incorporates the mixture structures to borrow information across features. Our method ranks the log fold change for the posterior samples after burn-in in the MCMC chains to get the top differentially abundant OTUs under two treatment groups in microbiome data. In this way, our method exceeds all other methods in the ranking of a differential abundance of OTUs across different conditions.
|