Inference of Significant Microbial Interactions From Longitudinal Metagenomics Data

Area of Science:

• Microbiology
• Computational Biology
• Systems Biology

Background:

• Next-generation sequencing (NGS) advances understanding of microbial ecosystems like the human gut.
• Inferring microbial interactions is challenging due to relative abundance data from 16S rDNA sequences, lacking absolute cell counts.

Purpose of the Study:

• To develop a procedure for describing microbial community dynamics and estimating interactions.
• To apply this procedure to gut microbiome data from children progressing to Type 1 diabetes (T1D) and compare with healthy controls.

Main Methods:

• Integration of generalized Lotka-Volterra equations, forward stepwise regression, and bootstrap aggregation.
• Validation using experimentally confirmed interactions in a cheese microbial community.
• Application to time-series 16S rDNA sequences of gut microbiomes.

Main Results:

• The number of inferred microbial interactions increased over the first three years of life.
• Healthy children exhibited more gut microbial interactions than T1D progressors.
• Specific interactions (e.g., inhibition of Bacteroidia by Actinobacteria/Bacilli) were conserved, while others (e.g., Gammaproteobacteria on Clostridia) differed between groups.

Conclusions:

• The developed procedure effectively infers microbial interactions from relative abundance data.
• Gut microbial communities in healthy children are more interactive than in T1D progressors.
• Distinct interaction patterns may contribute to T1D development, highlighting potential biomarkers.