Multimodal Metagenomic Analysis Reveals Microbial InDels as Superior Biomarkers for Pediatric Crohn’s Disease

The gut microbiome is closely associated with pediatric Crohn's disease (CD), while the multidimensional microbial signature and their capabilities for distinguishing pediatric CD are underexplored. This study aims to characterize the microbial alterations in pediatric CD and develop a robust classification model. A total of 1,175 fecal metagenomic sequencing samples, predominantly from three cohorts of pediatric CD patients, were re-analyzed from raw sequencing data using uniform process pipelines to obtain multidimensional microbial alterations in pediatric CD, including taxonomic profiles, functional profiles, and multi-type genetic variants. Random forest algorithms were used to construct classification models after comparing multiple machine learning algorithms. We found pediatric CD samples exhibited reduced microbial diversity and unique microbial characteristics. Pronounced abundance differences in 45 species and 1,357 KO genes. Particularly, Enterocloster bolteae emerged as a pivotal pediatric CD-associated species. Additionally, we identified a vast amount of microbial genetic variants linked to pediatric CD, including 192 structural variants, 1,256 insertions/deletions (InDels), and 3,567 single nucleotide variants, with a considerable portion of these variants occurred in non-genic regions. The InDel-based model outperformed other predictive models against multidimensional microbial signatures, achieving an AUC of 0.982. The robustness and disease specificity were further confirmed in an independent CD cohort (AUC=0.996) and five other microbiome-associated pediatric cohorts. Our study provided a comprehensive landscape of microbial alterations in pediatric CD and introduced a highly effective diagnostic model rooted in microbial InDels, which contributes to the development of the non-invasive diagnostic tools and targeted therapies.