Taxanorm: a novel taxa-specific normalization approach for microbiome data.

Area of Science:

• Microbiome analysis
• Bioinformatics
• Statistical modeling

Background:

• High-throughput sequencing studies exhibit variable sequencing depth across samples, complicating biological signal detection.
• Existing normalization methods often use sample-specific factors, leading to inaccurate corrections for certain taxa.
• Differential sequencing depth can obscure true biological signals and hinder inter-sample comparisons.

Purpose of the Study:

• To develop a novel normalization method, TaxaNorm, for microbiome data.
• To address limitations of existing methods by incorporating taxon-specific biases.
• To improve the accuracy of downstream analyses and data interpretation in microbiome studies.

Main Methods:

• Developed TaxaNorm, a normalization method utilizing a zero-inflated negative binomial model.
• Modeled sequencing depth effects on mean and dispersion as taxon-specific.
• Incorporated zero-inflation to better represent microbiome data characteristics.

Main Results:

• TaxaNorm demonstrates comparable performance to existing methods in simulations, with improved power in certain scenarios.
• The method effectively balances statistical power and false discovery control.
• Application to real-world data shows TaxaNorm enhances the correction of technical bias.

Conclusions:

• TaxaNorm effectively corrects for both sample- and taxon-specific biases in microbiome data using a regression framework.
• This normalization aids in clearer data interpretation and visualization.
• The TaxaNorm R package is publicly available for use.