A0281
Title: Analysis of microbiome differential abundance by pooling Tobit models
Authors: Gen Li - University of Michigan Ann Arbor (United States) [presenting]
Abstract: Microbiome differential abundance analysis (DAA) is pivotal in identifying microbial features associated with various disease conditions. However, the inherent complexities of metagenomics sequencing data, including compositionality and sparsity, challenge the accuracy and effectiveness of current methodologies, resulting in inflated false discovery rates and diminished statistical power. Addressing this gap, a novel approach called ADAPT (analysis of differential abundance by pooling Tobit models) is presented. Explicit assumptions are first established for DAA to elucidate the relationship between relative and absolute abundances. Leveraging this insight, ADAPT strategically identifies a subset of reference taxa based on the ordered list of fold changes in relative abundances. Subsequently, it conducts hypothesis testing on the fold change of count ratios between each taxon and the reference set. ADAPT employs Tobit models to effectively estimate fold changes, treating zero values as partially observed values left censored at the detection limit. Through extensive simulation studies and real data analysis, ADAPT is demonstrated to surpass existing methods by better controlling false discovery rates and exhibiting higher power.
