Microbial Co-occurrence Network Analysis

168 concepts134 papers1 proposal

Beyond the Microbial Parts List

Traditional microbiome profiling relies heavily on cataloging taxonomic diversity. By sequencing marker genes (like the 16S rRNA gene for bacteria or ITS region for fungi) or performing shotgun metagenomic sequencing, researchers typically generate a list of which organisms are present and their relative abundances. From there, standard metrics like alpha diversity (the richness within a single sample) and beta diversity (the differences in composition between samples) are computed.

However, these metrics treat microbes as isolated, independent entities. In reality, microbial communities are complex, interconnected ecosystems. Microbes constantly interact: they compete for limited nutrients, engage in syntrophic cross-feeding where one species' waste becomes another's food, or secrete antimicrobials to actively exclude competitors.

To capture this social landscape, researchers use microbial co-occurrence network analysis (also referred to as molecular ecological network analysis). This framework translates abstract abundance tables into structured mathematical graphs. By mapping how the presence and abundance of different organisms fluctuate together across multiple samples, this approach allows scientists to infer ecological relationships, identify critical "keystone" species, and quantify the overall stability of an ecosystem.

From Abundance Tables to Mathematical Graphs

The core intuition behind a co-occurrence network is simple: if two microbes are consistently found together, they may share an ecological niche or cooperate; if the presence of one reliably predicts the absence of another, they may compete or require mutually exclusive environmental conditions.

To build a network, researchers treat each microbial taxon (such as a genus or an operational taxonomic unit) as a node. An edge (a line connecting two nodes) is drawn if there is a statistically significant correlation between their abundance patterns across a dataset.

[Taxon A] --(Positive Correlation)--> [Taxon B]  (Co-occurrence / Mutualism)
[Taxon A] --(Negative Correlation)--> [Taxon C]  (Co-exclusion / Competition)

As highlighted by Jing Wang et al. in Sex differences in colonization of gut microbiota from a man with short-term vegetarian and inulin-supplemented diet in germ-free mice and single-author work by Qun Xu in How Does Extreme Drought Affect Phytoplankton Community Assembly in Aquatic Reserves? A Study from the Confluence of Poyang Lake and Yangtze River, China, standard correlation coefficients like Spearman’s rank or Pearson correlations serve as the initial mathematical foundation for mapping these relationships. A strong positive correlation denotes co-occurrence, while a strong negative correlation denotes co-exclusion, a pattern explicitly tracked by Yong Jiang et al. in Changes in Phyllosphere Microbial Communities of Pinus tabuliformis after Infestation by Bursaphelenchus xylophilus.

The Challenge of Compositional Data

Directly applying standard correlation formulas to sequencing data introduces a major mathematical artifact known as the compositionality problem. High-throughput sequencing measures relative abundance, not absolute abundance; the data must sum to 1 (or 100%).

If a single opportunistic bacterium blooms dramatically in a sample, the relative percentage of every other microbe automatically drops, purely due to the math of fractions. A standard correlation algorithm would interpret this zero-sum drop as a biological "co-exclusion" or competitive interaction, generating widespread false negatives.

To overcome this, modern network pipelines deploy specialized sparse correlation techniques designed for compositional data. As utilized by Qianpeng Zhang et al. in Accelerated dysbiosis of gut microbiota during aggravation of DSS-induced colitis by a butyrate-producing bacterium and Geicho Nakatsu et al. in Gut mucosal microbiome across stages of colorectal carcinogenesis, these algorithms mathematically transform relative abundances into log-ratios or apply sparse estimation frameworks. This ensures that an edge is only drawn when two microbes exhibit true biological coordination across samples, independent of total sequencing depth.

Decoding Network Topology

Once the statistical pipeline filters out spurious associations, the resulting graph can be analyzed using graph theory. Rather than evaluating microbes individually, researchers calculate topological metrics to understand the architecture of the entire community.

Hubs and Keystone Species

In any network, some nodes act as vital structural anchors. By measuring degree centrality (the total number of direct edges connected to a node) and betweenness centrality (how often a node acts as a bridge along the shortest path between other nodes), researchers can identify "keystone" or "hub" species.

Yan Kang et al. demonstrated this principle in Abundant fungi dominate the complexity of microbial networks in soil of contaminated site: High-precision community analysis by full-length sequencing, showing that removing highly connected fungal keystone nodes mathematically destabilizes the entire network structure. In plant ecosystems, single-author work by Heng Zheng in High-throughput sequencing-based analysis of the composition and diversity of endophytic bacteria community in tubers of Gastrodia elata f.glauca revealed that specific dominant phyla like Proteobacteria consistently emerge as high-connectivity structural hubs, maintaining the organization of internal plant tissues.

Modularity and Functional Clusters

Microbial networks are rarely uniform grids; they naturally segregate into modules—tightly knit clusters of nodes that interact heavily with each other but sparsely with the rest of the network. These modules often represent functional subcommunities or shared ecological niches. Ziliang Zhu et al. leveraged community detection algorithms in Gut microbiota maturation and early behavioral and cognitive development to isolate these functional modules within the infant gut, tracking how the emergence of specific structural clusters correlates with host cognitive milestones.

Cooperative vs. Competitive Dynamics

The nature of the edges themselves provides a proxy for community behavior. By calculating the ratio of positive to negative edges, researchers can infer whether an ecosystem is governed by cooperative relationships (like cross-feeding) or competitive exclusion. Single-author work by Zhibo Ma in Effects of Facility Cultivation Shaping Soil Microbial Community Structure in Jujube Orchard demonstrated that tracking this edge ratio reveals how external agricultural management shifts the underlying ecological balance between cooperation and competition.

Quantifying Stability and Stress Responses

The true power of co-occurrence networks lies in their ability to measure ecosystem resilience. A network's topology directly reflects its capacity to withstand environmental stress or disease.

A highly complex network—characterized by a high average clustering coefficient and dense connectivity—indicates a mature, stable community where metabolic tasks are distributed efficiently, as detailed by Kun Cheng et al. in Interaction between dissolved organic carbon and fungal network governs carbon mineralization in paddy soil under co-incorporation of green manure and biochar. When an ecosystem is healthy, its network is robust; even if a few species are suppressed, alternative pathways and redundant connections prevent the community from collapsing.

[Healthy Ecosystem] -> Dense, highly connected network -> High resilience
[Stressed Ecosystem] -> Fragmented, simplified network -> High fragility

When a system undergoes a severe perturbation, the network topology fractures. For example, Megha Singhal et al. showed in Serotonin Transporter Deficiency is Associated with Dysbiosis and Changes in Metabolic Function of the Mouse Intestinal Microbiome how genetic host deficiencies can fragment the gut microbial network structure, reducing its connectivity.

By comparing network properties across different conditions, researchers can diagnose how stressors simplify and weaken communities. This pattern of network simplification under stress has been documented across diverse environments, including chemical flooding in oil reservoirs studied by Guoling Ren et al. in Compositions and Co-occurrence Patterns of Bacterial Communities Associated With Polymer- and ASP-Flooded Petroleum Reservoir Blocks, and water scarcity in crops analyzed by Sasa Xie et al. in Drought Sensitivity of Sugarcane Cultivars Shapes Rhizosphere Bacterial Community Patterns in Response to Water Stress.

Expanding the Interactome Across Domains

The mathematical abstraction of co-occurrence analysis means it is not restricted to looking at bacteria in isolation. Modern applications construct multi-domain and multi-omic networks to map the broader "interactome."

Inter-Kingdom Networks

Ecosystems contain multiple kingdoms of life operating simultaneously. Researchers use co-occurrence modeling to bridge these domains, mapping bacteria alongside fungi or viruses. Olabisi Oluwabukola Coker et al. applied this in Enteric fungal microbiota dysbiosis and ecological alterations in colorectal cancer to uncover how bacterial-fungal relationships shift from cooperation to competition during cancer progression.

Similarly, Mo Li et al. in More Positive or More Negative? Metagenomic Analysis Reveals Roles of Virome in Human Disease-Related Gut Microbiome and Geoffrey D. Hannigan et al. in Biogeography and environmental conditions shape bacteriophage-bacteria networks across the human microbiome utilized graph theory to model phage-bacteria networks, identifying specific viruses that act as structural bridges to maintain the stability of the surrounding bacterial community.

Multi-Omic Integration

Beyond mapping organisms to other organisms, edges can connect microbial taxa to non-biological variables, such as soil chemistry, host metabolites, or clinical phenotypes.

As demonstrated by Shi Qi et al. in Chlorine Dioxide Reprograms Rhizosphere Microbial Communities to Enrich Interactions with Tobacco (Nicotiana tabacum) and Yang Cao et al. in Dopamine alleviates cadmium stress in apple trees by recruiting beneficial microorganisms to enhance the physiological resilience revealed by high-throughput sequencing and soil metabolomics, integrating sequencing data with metabolomic profiles into a single unified network allows researchers to trace the functional consequences of microbial shifts. These multi-omic networks reveal not just who is present, but how the collective microbial web drives the chemical and physical traits of the host organism.

Key Foundational Work

Molecular ecological network analyses (2012) — contributed 3 concept elements to this direction.
History & citationVersion 2 · 2026-10-04

Current · Permanent link

  1. Version 2 (latest) · 2026-10-04 · 132 papers
  2. Version 1 · 2026-10-04 · 89 papers
Cite this direction

Latest version (a living page):

@misc{lacuna_co_occurrence_network_analysis_0604b37c,
  title = {co-occurrence network analysis},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/co-occurrence-network-analysis/rdr_0604b37ca38242009a9b72b06c365046},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 2, frozen text):

@misc{lacuna_co_occurrence_network_analysis_0604b37c_v2,
  title = {co-occurrence network analysis},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/co-occurrence-network-analysis/rdr_0604b37ca38242009a9b72b06c365046/v/2},
  note = {Version 2, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake