NBZIMM: negative binomial and zero-inflated mixed models, with application to microbiome/metagenomics data analysis.

Area of Science:

• Microbiology
• Bioinformatics
• Statistical Genetics

Background:

• Microbiome and metagenomic datasets exhibit unique characteristics like variable read counts, over-dispersion, and zero-inflation.
• Longitudinal study designs in microbiome research introduce further analytical complexities.
• Existing analytical tools often lack the flexibility required for processed multilevel or longitudinal microbiome/metagenomic data.

Purpose of the Study:

• To develop a flexible and efficient R package for analyzing processed multilevel or longitudinal microbiome/metagenomic data.
• To provide statistical methods and implementations specifically addressing the unique data characteristics and complex designs in microbiome/metagenomic studies.

Main Methods:

• The NBZIMM R package implements negative binomial mixed models, zero-inflated negative binomial mixed models, and zero-inflated Gaussian mixed models.
• Functions for numerical and graphical summarization of fitted model results are included.
• Core functionalities leverage established R packages (nlme, MASS) for robust analysis of over-dispersed and zero-inflated count or proportion data with multilevel structures.

Main Results:

• NBZIMM facilitates the analysis of complex microbiome/metagenomic data with multilevel structures, such as longitudinal studies.
• The package provides a unified framework for handling over-dispersion and zero-inflation common in these datasets.
• It enables both numerical and graphical summarization of model outputs.

Conclusions:

• The NBZIMM package offers valuable tools for the sophisticated analysis of microbiome and metagenomic data.
• It addresses key challenges posed by data characteristics and complex study designs in the field.