MNRS: Multi-Factor Network-Based Ranking Score for Detecting Critical Transitions of Complex Diseases Using Gut Microbial Data

Disease progression is not always gradual and may instead involve abrupt deterioration, with a critical threshold separating pre-deterioration and post-deterioration states. Detecting such pre-disease states is of major importance because they often precede catastrophic transitions. Increasing evidence suggests that the onset and progression of many diseases, including type 1 diabetes, celiac disease, and colorectal cancer, are closely associated with the gut microbiome. Although transcriptome-based approaches, particularly those relying on gene expression data, have been widely used to identify critical states in biological systems, they are often not well suited to gut microbiome data because of its sparsity, compositionality, and substantial noise. Here, we propose a novel computational framework, termed multi-factor network-based ranking score (MNRS), for detecting pre-disease states from gut microbiome data. MNRS infers perturbed microbial networks and quantifies dynamic alterations in species- or genus-level associations, thereby enabling the detection of early-warning signatures of critical transitions. Analyses of both simulated data and multiple real-world datasets show that MNRS accurately identifies pre-disease states and outperforms existing methods in both robustness and detection performance. In addition, MNRS reveals sensitive “dark species” overlooked by conventional differential abundance analyses but potentially important in disease deterioration.