Quantitative Amplicon Sequencing Is Necessary to Identify Differential Taxa and Correlated Taxa Where Population Sizes Differ

Abstract

High-throughput, multiplexed-amplicon sequencing has become a core tool for understanding environmental microbiomes. As researchers have widely adopted sequencing, many open-source analysis pipelines have been developed to compare microbiomes using compositional analysis frameworks. However, there is increasing evidence that compositional analyses do not provide the information necessary to accurately interpret many community assembly processes. This is especially true when there are large gradients that drive distinct community assembly processes. Recently, sequencing has been combined with Q-PCR (among other sources of total quantitation) to generate "Quantitative Sequencing" (QSeq) data. QSeq more accurately estimates the true abundance of taxa, is a more reliable basis for inferring correlation, and, ultimately, can be more reliably related to environmental data to infer community assembly processes. In this paper, we use a combination of published data sets, synthesis, and empirical modeling to offer guidance for which contexts QSeq is advantageous. As little as 5% variation in total abundance among experimental groups resulted in more accurate inference by QSeq than compositional methods. Compositional methods for differential abundance and correlation unreliably detected patterns in abundance and covariance when there was greater than 20% variation in total abundance among experimental groups. Whether QSeq performs better for beta diversity analysis depends on the question being asked, and the analytic strategy (e.g., what distance metric is being used); for many questions and methods, QSeq and compositional analysis are equivalent for beta diversity analysis. QSeq is especially useful for taxon-specific analysis; QSeq transformation and analysis should be the default for answering taxon-specific questions of amplicon sequence data. Publicly available bioinformatics pipelines should incorporate support for QSeq transformation and analysis.

Keywords: Compositional data; Data transformation; Differential abundance; Microbiome; QSeq.