gFlora: A Topology-Aware Method to Discover Functional Co-Response Groups in Soil Microbial Communities

Area of Science:

• Soil microbiology
• Ecology
• Bioinformatics

Background:

• Soil microorganisms drive essential ecosystem processes like nutrient cycling.
• Understanding the link between specific microbial taxa and soil functions is crucial for soil management but is complex.
• Existing methods struggle with the scale and complexity of soil microbial communities.

Purpose of the Study:

• To develop a method for identifying functional co-response groups of soil taxa.
• To model soil microbial communities as ecological co-occurrence networks.
• To leverage graph convolution for discovering taxon-to-function links.

Main Methods:

• Modeled soil microbial communities as weighted co-occurrence networks.
• Applied graph convolution (gFlora) to compute group co-response effects.
• Evaluated gFlora on four real-world soil microbiome datasets.

Main Results:

• gFlora outperformed competing methods across all evaluation metrics.
• The method identified new functional roles for under-studied microbial taxa.
• Graph convolution proved vital for including low-abundance taxa, reducing bias.

Conclusions:

• gFlora effectively identifies functional links between microbial taxa and soil functions.
• The approach reveals that diverse microbial genera form tightly connected groups, playing collaborative roles.
• This network-based method enhances our ability to understand and manage soil ecosystems.