A joint modeling approach for longitudinal microbiome data improves ability to detect microbiome associations with...

Area of Science:

• Microbiology
• Biostatistics
• Computational Biology

Background:

• Microbiome composition changes are linked to human diseases.
• Quantifying these changes' association with disease onset is challenging due to limited analytical techniques.
• Existing joint modeling approaches struggle with non-Gaussian microbiome data.

Purpose of the Study:

• To present a novel joint modeling framework for longitudinal microbiome data.
• To incorporate time-dependent microbial abundances into time-to-event analyses.
• To provide a statistically robust method for microbiome-disease association studies.

Main Methods:

• Developed a joint modeling framework using a negative binomial mixed-effects model for longitudinal taxon abundances.
• Integrated modeled microbial abundances into a hazard function with a parameterization suitable for proportional microbiome data.
• Validated the approach through simulations and analysis of a pregnancy microbiome dataset.

Main Results:

• The proposed joint modeling framework demonstrated improved performance over existing methods.
• The method successfully identified associations between longitudinal microbial changes and disease onset.
• Biologically interpretable results were generated, accounting for the proportional nature of microbiome data.

Conclusions:

• The joint modeling framework offers a powerful methodology for analyzing longitudinal microbiome count data.
• This approach can uncover significant associations between microbial composition dynamics and disease development.
• The findings have the potential to inform the development of diagnostic biomarkers and clinical interventions.