An expectation-maximization algorithm enables accurate ecological modeling using longitudinal microbiome sequencing...

Area of Science:

• Microbiology
• Computational Biology
• Ecology

Background:

• Microbial community dynamics are driven by complex interactions, many of which remain poorly understood.
• High-throughput microbiome data allows for ecological model inference, but is limited by the lack of absolute cell density measurements (biomass).

Purpose of the Study:

• To present a novel computational approach, BEEM (Biomass Estimation and Model Inference), that resolves the limitation of biomass data in ecological model inference.
• To enable the construction of accurate ecological models of microbial communities from high-throughput sequencing data.

Main Methods:

• Developed an expectation-maximization algorithm (BEEM) that couples biomass estimation with generalized Lotka-Volterra model (gLVM) inference.
• Applied BEEM to public microbiome datasets lacking biomass information.

Main Results:

• BEEM outperforms existing state-of-the-art methods for inferring gLVMs.
• BEEM successfully constructed personalized ecological models of human gut microbial communities without requiring experimental biomass data.
• Identified personalized dynamics and keystone species within individual microbiomes.

Conclusions:

• BEEM overcomes a critical bottleneck in microbiome systems analysis.
• Enables accurate ecological model inference from high-throughput sequencing data without experimental biomass measurements.
• Facilitates deeper understanding of microbiome structure and function.