An Omnibus Test for Differential Abundance Analysis of Metagenomic Data

• Activity Number: 420
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 2:00 PM to 3:50 PM
• Sponsor: ENAR
• Author(s): Jun Chen* and Diane Grill and Karla Ballman
• Companies: Mayo Clinic and Mayo Clinic and Mayo Clinic
• Keywords: Statistical genetics ; Metagenomics ; Genomics ; Next generation sequencing ; Zeroinflated count data ; Statistical genomics

Abstract

One central goal of metagenomic studies is to identify taxa that show differentiation between sample groups. The identified taxa can be used as biomarkers for disease diagnosis and prevention. Many methods have been developed to address this problem ranging from simple adaptation of the t test (Metastat) to more sophisticated statistical test based on zero-inflated Gaussian model (metagenomeSeq) and count based methods (DESeq2). However, none of the statistical methods have taken into account all the features of the taxa data, which are zero-inflated overdispersed count data. Moreover, most of the methods focus on detecting the change of the mean of the taxa abundance. In real situation, disease could affect not only the abundance mean but also the prevalence and the variance. Both dysbiosis and disease heterogeneity can lead to differential variance. We therefore develop an omnibus test based on a zero-inflated count model that jointly tests the equality of mean, zero probability and variance between sample groups. Both simulations and real data applications demonstrated the increased power of the omnibus test as well better control of the type I error than existing methods.