Comprehensive microbiota alterations in IBD and improved diagnostic accuracy for IBD using multi‐kingdom microbial features.

R3269 --> Background: Inflammatory bowel disease(IBD),including Crohn's Disease(CD) and ulcerative colitis (UC), is closely linked with altered gut microbiota.Roles of bacteria in IBD have been widely investigated, whereas other microbial kingdoms, fungi, archaea, virus, are less explored. Here, we aim to investigate the multi‐kingdom microbial alterations and their capabilities for IBD diagnosis. Methods: A total of 968 stool samples (292 CD, 305 UC and 371 controls) of untreated patients from iHMP were collected. Differential microorganisms were identified via wilcoxon test followed by adjustment for confounders. Inter‐kingdom co‐occurrence interactions analysis was performed by SparCC then microorganism modules were clustered by "MCODE". Subsequently, neural networks algorithm was used to build diagnostic models for CD or UC with single‐ or multi‐kingdom microbial features and the best panel of biomarkers was further identified by recursive feature elimination. Moreover, diagnostic models were validated in additional independent cohorts. Differential microbial pathways were analyzed to further explore functional alterations in CD and UC. Results: In CD and UC, the microbial compositional change was identified in different kingdoms. Compared to controls, we identified 41 bacteria, 37 fungi, 40 archaea and 50 viruses with differential abundance in CD. Similarly, 63 bacteria, 52 fungi, 58 archaea and 39 viruses with differential abundance were detected in UC. Moreover, inter‐kingdom interactions were more complex in IBD than controls, especially the interactions between fungal Aspergillusand BeAn 58058 virus. Notably, the diagnostic models constructed with multi‐kingdom microbial features performed better (AUC of 0.90 in CD and 0.91 in UC) than that with single kingdom microbial features. The diagnostic models also showed reliable accuracy in independent cohorts with an average AUC of 0.93 (CD) and 0.74 (UC), respectively. Among these biomarkers, bacterial Bacteroidesand fungalAspergillusare the main contributors to the distinguishing ability in both CD and UC diagnostic models. In addition, pathways related to lipid and terpenoids metabolism and immune response were altered in CD and UC, such as glycosphingolipid biosynthesis, carotenoid biosynthesis and IL‐17 signaling pathway. Conclusions: Our study investigated comprehensivepathogenic microbiota of IBD, which encompasses bacteria, fungi, virus and archaea. And the combination of multi‐kingdom microbial biomarkers may serve as effective tools for CD and UC diagnosis, which outperforms single kingdom biomarkers. The alterations of microbial inter‐kingdom interactions and cross‐kingdom functional changes provide new insight in understanding pathological mechanisms of IBD. [ABSTRACT FROM AUTHOR]