From variability to stability: Sensitivity of network properties in IBD human gut microbiome studies

Theresa Geese; Astrid Dempfle

doi:10.1016/j.csbj.2025.05.005

From variability to stability: Sensitivity of network properties in IBD human gut microbiome studies

Comput Struct Biotechnol J. 2025 May 10:27:1945-1961. doi: 10.1016/j.csbj.2025.05.005. eCollection 2025.

Authors

Theresa Geese¹, Astrid Dempfle¹

Affiliation

¹ Institute of Medical Informatics and Statistics, Kiel University and University Hospital Schleswig-Holstein, Kiel, Germany.

Abstract

Background: The gut microbiome's role in inflammatory bowel disease (IBD) is well-established, but capturing its complexity is challenging. Network analysis offers a valuable approach, but selecting suitable measures is crucial. This study examines the sensitivity of network properties to abundance variations. It evaluates whether these properties reflect the microbiome in IBD or are too sensitive to variability from e.g. laboratory conditions or intra-individual changes.

Methods: Using genetically unrelated individuals from the KINDRED cohort (IBD n = 522, healthy controls n = 365) and the PRISM cohort (IBD n = 42, healthy controls n = 42), microbial networks were constructed with genera as nodes and significant pairwise correlations as edges, separately for IBD patients and controls. Important IBD-related nodes, identified through centrality measures, and non-disease-related nodes were varied in abundance ( ± 30 %), and networks were re-constructed and compared with initial networks regarding local and global properties.

Results: Network properties in IBD were sensitive to abundance variations, with small and large changes producing similar effects. Sensitivity to increasing read counts of disease-related and non-disease-related genera was similar. Local properties showed magnitude-dependent changes of up to 50 % in response to the depletion of disease-related genera, relative to no modification applied, and an almost binary sensitivity pattern when modifying non-disease-related genera. Global case network properties changed less than 10 % in most settings, potentially indicating a certain stability of dysbiosis.

Conclusion: Caution is needed with network-based approaches, as even small variations, stemming from sources of microbiome variability, can affect results and reproducibility. The relatively stable dysbiosis in IBD could pose challenges for microbiome-directed therapies.

Keywords: Gut microbiome; Inflammatory bowel disease; Network analysis; Network properties; Sensitivity; Varying microbiome abundance.