Altered gut bacterial and metabolic signatures and their interaction in inflammatory bowel disease

Abstract

Dysregulation of the gut microbiome has been implicated in the progression of many diseases. This study explored the role of microbial and metabolic signatures, and their interaction between the Human inflammatory bowel disease (IBD) and healthy controls (HCs) based on the combination of machine learning and traditional statistical analysis, using data collected from the Human Microbiome Project (HMP) and the Integrative Human Microbiome Project (iHMP). It was showed that the microbial and metabolic signatures of IBD patients were significantly different from those of HCs. Compared to HCs, IBD subjects were characterized by 25 enriched species and 6 depleted species. Furthermore, a total of 17 discriminative pathways were identified between the IBD and HC groups. Those differential pathways were mainly involved in amino acid, nucleotide biosynthesis, and carbohydrate degradation. Notably, co-occurrence network analysis revealed that non-predominant bacteria Ruminococcus_obeum and predominant bacteria Faecalibacterium_prausnitzii formed the same broad and strong co-occurring relationships with pathways. Moreover, the essay identified a combinatorial marker panel that could distinguish IBD from HCs. Receiver Operating Characteristic (ROC) and Decision Curve Analysis (DCA) confirmed the high accuracy (AUC = 0.966) and effectiveness of the model. Meanwhile, an independent cohort used for external validation also showed the identical high efficacy (AUC = 0.835). These findings showed that the gut microbes may be relevant to the pathogenesis and pathophysiology, and offer universal utility as a non-invasive diagnostic test in IBD.

Keywords

• Inflammatory bowel disease
• Gut microbiome
• Metabolic pathways
• Machine learning
• Diagnosis