Cross-biome comparison of microbial association networks

Karoline Faust, Gipsi Lima‐Mendez, Jean‐Sébastien Lerat, J. Fah Sathirapongsasuti, Rob Knight, Curtis Huttenhower, Tom Lenaerts, Jeroen RaesFrontiers in Microbiology 2015 · DOI

Abstract

Clinical and environmental meta-omics studies are accumulating an ever-growing amount of microbial abundance data over a wide range of ecosystems. With a sufficiently large sample number, these microbial communities can be explored by constructing and analyzing co-occurrence networks, which detect taxon associations from abundance data and can give insights into community structure. Here, we investigate how co-occurrence networks differ across biomes and which other factors influence their properties. For this, we inferred microbial association networks from 20 different 16S rDNA sequencing data sets and observed that soil microbial networks harbor proportionally fewer positive associations and are less densely interconnected than host-associated networks. After excluding sample number, sequencing depth and beta-diversity as possible drivers, we found a negative correlation between community evenness and positive edge percentage. This correlation likely results from a skewed distribution of negative interactions, which take place preferentially between less prevalent taxa. Overall, our results suggest an under-appreciated role of evenness in shaping microbial association networks.

Introduction

Microbes do not live in isolation; they form complex webs of competition, cooperation, and cross-feeding. Researchers use "co-occurrence networks"—built from 16S rDNA sequencing data—to infer these relationships. However, most studies focus on a single environment, such as the human gut or a specific plot of land. This study, published in Frontiers in Microbiology, provides a large-scale meta-analysis to determine if universal rules govern microbial networks across diverse biomes or if observed differences are merely artifacts of technical biases like sequencing depth or sample size.

Methods and Datasets

The authors analyzed 20 different 16S rDNA sequencing datasets drawn from major repositories, including the QIIME Database, the Human Microbiome Project (HMP), and the Earth Microbiome Project (EMP). These covered biomes ranging from the human intestine and skin to tundra and grasslands.

To construct the networks, the researchers employed a sophisticated bioinformatic pipeline called CoNet. Unlike standard methods that rely on a single correlation metric, CoNet uses an ensemble approach, combining correlation measures (Pearson and Spearman) with similarity/dissimilarity measures (Bray-Curtis and Kullback-Leibler). To handle the "compositional" nature of sequencing data, they used the ReBoot procedure to calculate significance and performed extensive Dirichlet-Multinomial simulations to test how data properties like evenness affect network construction.

Key Insights: The Role of Evenness

The study revealed a clear dichotomy between soil and host-associated environments. Soil networks are generally "thinner" (lower density) and "colder" (lower percentage of positive edges) than host-associated networks.

Comparison of microbial association networks across different biomes: (A) soil, (B) gut, (C) forest, and (D) skin. The legend indicates phylum color codes and interaction types (positive vs. negative).

A primary driver of these differences is community evenness—how equally distributed species abundances are. Soil communities are generally very even, while host communities are often dominated by a few highly abundant taxa. The authors propose that negative interactions (competition) are likely present in all biomes but are harder to detect in uneven communities. In the human gut, the "losers" of competition are often driven to such low abundances that they fall below the detection limit of the sequencer. In the even environment of soil, these rare taxa remain visible, allowing their negative correlations to be statistically detected.

Key Results

  • Positive Edge Percentage (PEP): Soil networks had significantly lower PEP than host networks ($p = 0.0002$).
  • Evenness Correlation: There was a strong negative correlation ($R^2 = 0.69$) between Sheldon evenness and PEP. As a community becomes more even, the percentage of positive edges drops.
  • Technical Drivers: After rigorous testing, the authors concluded that sample number and sequencing depth were not the primary reasons for the differences between soil and host networks.

Limitations and Future Directions

While co-occurrence networks are powerful, they detect associations rather than direct physical interactions; two microbes might co-occur simply because they share environmental preferences. Furthermore, 16S data often lacks the resolution to distinguish between closely related strains with different ecological roles. Future work focusing on longitudinal, time-series data will be essential to see how these networks shift dynamically over time.

Summary updated

Abstract

Clinical and environmental meta-omics studies are accumulating an ever-growing amount of microbial abundance data over a wide range of ecosystems. With a sufficiently large sample number, these microbial communities can be explored by constructing and analyzing co-occurrence networks, which detect taxon associations from abundance data and can give insights into community structure. Here, we investigate how co-occurrence networks differ across biomes and which other factors influence their properties. For this, we inferred microbial association networks from 20 different 16S rDNA sequencing data sets and observed that soil microbial networks harbor proportionally fewer positive associations and are less densely interconnected than host-associated networks. After excluding sample number, sequencing depth and beta-diversity as possible drivers, we found a negative correlation between community evenness and positive edge percentage. This correlation likely results from a skewed distribution of negative interactions, which take place preferentially between less prevalent taxa. Overall, our results suggest an under-appreciated role of evenness in shaping microbial association networks.
BibTeX
@inproceedings{faust2015cross,
  author = {Karoline Faust and Gipsi Lima‐Mendez and Jean‐Sébastien Lerat and J. Fah Sathirapongsasuti and Rob Knight and Curtis Huttenhower and Tom Lenaerts and Jeroen Raes},
  title = {{Cross-biome comparison of microbial association networks}},
  booktitle = {Frontiers in Microbiology},
  year = {2015},
  doi = {10.3389/fmicb.2015.01200},
}
Figure