BZINB Model-Based Pathway Analysis and Module Identification Facilitates Integration of Microbiome and Metabolome...

Area of Science:

• Microbiology
• Bioinformatics
• Systems Biology

Background:

• Integrating multi-omics data (e.g., microbiome and metabolome) is crucial for understanding human health and disease.
• Existing correlation-based network analyses struggle with the excess zeros common in microbiome data.

Purpose of the Study:

• To introduce a novel bivariate zero-inflated negative binomial (BZINB) model for improved microbiome-metabolome data integration.
• To develop a network and module analysis method (BZINB-iMMPath) that accommodates excess zeros in multi-omics data.

Main Methods:

• Developed and applied a bivariate zero-inflated negative binomial (BZINB) model.
• Constructed metabolite-species and species-species correlation networks using BZINB.
• Identified correlated species modules using BZINB and similarity-based clustering.
• Utilized data from the ZOE 2.0 childhood oral health study.

Main Results:

• The BZINB model demonstrated superior accuracy over Spearman and Pearson correlations for microbiome-metabolome data.
• BZINB-iMMPath successfully constructed correlation networks and identified modules.
• Analysis of ZOE 2.0 data revealed differences in microbial-metabolite correlations associated with early childhood caries (ECC).

Conclusions:

• The BZINB model is a robust alternative for analyzing zero-inflated bivariate count data in multi-omics studies.
• BZINB-iMMPath enhances the integration of microbiome and metabolome data, offering insights into disease mechanisms.
• This approach is valuable for studies involving microbiome and metabolome data, such as those investigating oral health.