Dissecting the Ecological Structure of Health and Disease in the Global Gut Microbiome

The gut microbiota plays a crucial role in human health, but its coordinated ecological dynamics remain largely unclear. We present Wiredancer, a novel scalable framework based on similarity-constrained non-negative matrix factorization (NMF), which extracts continuous and overlapping microbial ecological factors (MEFs). By integrating 20,178 metagenomes spanning 36 countries and over 50 disease states, Wiredancer identified three robust and interpretable MEFs delineating the health-disease continuum. MEF1, the dysbiotic factor dominated by Bacteroides uniformis, was elevated in disease populations; MEF2, the protective factor characterized by Prevotella copri, was reduced compared with the healthy group; and MEF3, the intermediate factor represented by Bifidobacterium adolescentis, reflected a mixed ecological configuration between MEF1 and MEF2. MEFs exhibited high reproducibility across individuals and longitudinal cohorts, but showed significantly increased variability in disease, consistent with the Anna Karenina principle and highlighting disrupted ecological stability. These findings were validated in the largest Chinese metagenomic cohort of major psychiatric disorders, where MEFs were associated with clinical symptoms, peripheral biomarkers, and disease subtypes, and remained essentially stable under short-term treatment. Together, Wiredancer provides a generalizable strategy to define microbiome states and decode ecological transitions, offering new opportunities for precision diagnostics and stratified medicine in complex disorders.