Analysis and correction of compositional bias in sparse sequencing count data

Abstract

BackgroundCount data derived from high-throughput deoxy-ribonucliec acid (DNA) sequencing is frequently used in quantitative molecular assays. Due to properties inherent to the sequencing process, unnormalized count data is compositional, measuring relative and not absolute abundances of the assayed features. This compositional bias confounds inference of absolute abundances. Commonly used count data normalization approaches like library size scaling/rarefaction/subsampling cannot correct for compositional or any other relevant technical bias that is uncorrelated with library size.ResultsWe demonstrate that existing techniques for estimating compositional bias fail with sparse metagenomic 16S count data and propose an empirical Bayes normalization approach to overcome this problem. In addition, we clarify the assumptions underlying frequently used scaling normalization methods in light of compositional bias, including scaling methods that were not designed directly to address it.ConclusionsCompositional bias, induced by the sequencing machine, confounds inferences of absolute abundances. We present a normalization technique for compositional bias correction in sparse sequencing count data, and demonstrate its improved performance in metagenomic 16s survey data. Based on the distribution of technical bias estimates arising from several publicly available large scale 16s count datasets, we argue that detailed experiments specifically addressing the influence of compositional bias in metagenomics are needed.Electronic supplementary materialThe online version of this article (10.1186/s12864-018-5160-5) contains supplementary material, which is available to authorized users.

Absolute quantitation of microbes using 16S rRNA gene metabarcoding: A rapid normalization of relative abundances by quantitative PCR targeting a 16S rRNA gene spike‐in standard

“…A variety of pipelines is available to then get a relative abundance table of operational taxonomic units (OTUs; Sun et al, 2011). While the choice of the pipeline can certainly affect the results, the training set (Werner et al, 2012) and the method of normalization (Kumar et al, 2018) also have a major impact even though it is seldom discussed. The most common normalization procedure consists in dividing by the total number of reads in order to obtain the proportion of each OTU.…”

Assessment of statistical methods from single cell, bulk RNA-seq and metagenomics applied to microbiome data

“…Secondly, some normalizations, such as the geometric mean method implemented in DESeq2 or the trimmed mean of M-values of edgeR, have size factors mathematically equivalent or very similar to the compositional log-ratio proposed by Aitchison 24,36 . This has been shown to reduce the impact of compositionality on DA results 37 . We did not test the ANCOM package 38 because it was too slow for assessment in the simulation studies.…”

Disentangling microbial associations from hidden environmental and technical factors via latent graphical models

“…Important features include the level of excess zeros in the samples as well as the sequencing platform used ( Figure 3B-D). The strong influence of the number of zeros in the samples may be alleviated by improved scale estimations prior to data transformations [29] or zero-aware rank-based transformations [54]. However, we also recognized that many inferred latent components for the AGP data did not correlate with any of the measured covariates.…”