A review of normalization and differential abundance methods for microbiome counts data

Abstract

The recent development of cost‐effective high‐throughput DNA sequencing technologies has tremendously increased microbiome research. However, it has been well documented that the observed microbiome data suffers from compositionality, sparsity, and high variability. All of which pose serious challenges when analyzing microbiome data. Over the last decade, there has been considerable amount of interest into statistical and computational methods to tackle these challenges. The choice of inference aids in the selection of the appropriate statistical methods since only a few methods allow inferences for absolute abundance while most methods allow inferences for relative abundances. An overview of recent methods for differential abundance analysis and normalization of microbiome data is presented, focusing on methods that are accessible but have not been widely covered in previous literature. In detailed descriptions of each method, we discuss assumptions and if and how these methods address the challenges of microbiome data. These methods are compared based on accuracy metrics in real and simulated settings. The goal is to provide a comprehensive but non‐exhaustive set of potential and easily‐accessible tools for differential abundance and normalization of microbiome data.