Structured Abstract
INTRODUCTION
The human gut microbiome is a complex ecological system crucial for host health. Dysbiosis—i.e., the imbalance of gut microbial communities—is associated with a wide range of diseases, including obesity, diabetes, inflammatory bowel disease (IBD), Clostridioides difficile infection (CDI), irritable bowel syndrome (IBS), and colorectal cancer (CRC). Therapeutic approaches, such as fecal microbiota transplantation, dietary interventions, and probiotics, aim to restore balance by reshaping community composition. However, their outcomes remain inconsistent and unpredictable, in part owing to our limited understanding of the metabolic and ecological interactions that govern microbiome dynamics. The latter has also prevented the development of robust biomarkers to distinguish health from disease.
RATIONALE
Previous studies have suggested that health and dysbiosis may represent alternative community states but have not provided a mechanistic justification. Most efforts to define dysbiosis aim to identify bacterial taxa or functions that may differ between healthy and diseased communities or assume that reduced diversity is a universal hallmark of disease. However, such signatures vary across conditions and cohorts and fail to capture the ecological principles that shape disease states. To provide a more mechanistic understanding of gut microbial dynamics in health and disease, we developed a metabolically explicit model in which bacterial interactions arise naturally from competition for shared resources and cross-feeding.
RESULTS
The model reproduces key macroecological patterns and captures the functional redundancy characteristic of real gut microbiomes. Moreover, our model revealed the emergence of two distinct ecological states (healthy and dysbiotic states) whose α and β diversities, dominance indices, and numbers of functions and excreted metabolites closely resembled those observed in real microbiomes. The healthy state was dominated by competitive interactions, whereas the dysbiotic state was shaped by tightly connected cross-feeding consortia.
We also developed the ecological network balance index (ENBI), a metric that measures the relative contribution of positive versus negative interactions and reliably separates healthy from dysbiotic states. Calculating the ENBI for the model and metagenomic data for IBD, IBS, CDI, and CRC showed that, in all cases, diseased microbiomes exhibited higher ENBI values. Our metric also correlated with disease stage. These results proved robust across subsampling, geography, taxonomic levels, and profiling methods.
CONCLUSION
Unlike diversity-based metrics, which vary across diseases and cohorts, the ENBI consistently distinguishes healthy from diseased states and even tracks disease progression, offering a path toward robust, noninvasive early warning indicators of disease.
The ENBI also provides mechanistic insights: Our results show that dysbiosis is associated with a shift in the community interaction network, with positive interactions increasingly dominating over negative ones. Our framework is both general and extensible, and thus it can be adapted to incorporate additional biological features for the study of specific gut phenomena and be readily applied to other microbiomes (from the vaginal and oral to plant and soil ecosystems), or it can be used for the study and prediction of potential outcomes in therapeutic interventions. By linking microbial ecology with clinical research, our framework advances precision medicine and supports the development of more personalized strategies to maintain or restore gut health.
