Analysis and correction of compositional bias in sparse sequencing count data.

Area of Science:

• Genomics
• Bioinformatics
• Metagenomics

Background:

• High-throughput DNA sequencing generates count data used in molecular assays.
• Unnormalized count data exhibits compositional bias due to sequencing properties, hindering absolute abundance inference.
• Standard normalization methods like library size scaling do not correct for compositional bias.

Purpose of the Study:

• To address the limitations of existing methods in estimating compositional bias for sparse metagenomic 16S count data.
• To propose and validate a novel empirical Bayes normalization approach for compositional bias correction.
• To clarify the assumptions of common scaling normalization methods regarding compositional bias.

Main Methods:

• Demonstration of failure of existing compositional bias estimation techniques with sparse metagenomic 16S data.
• Development and application of an empirical Bayes normalization approach.
• Analysis of assumptions underlying scaling normalization methods.

Main Results:

• Existing techniques fail to accurately estimate compositional bias in sparse metagenomic 16S data.
• The proposed empirical Bayes method effectively corrects compositional bias.
• Clarification of assumptions for various scaling normalization methods.

Conclusions:

• Sequencing-induced compositional bias confounds absolute abundance inference.
• A new normalization technique improves compositional bias correction in sparse sequencing data, particularly for metagenomic 16S surveys.
• Further research is needed to fully understand compositional bias in metagenomics.