Multiomic analysis of Crohn's disease and Ulcerative colitis using machine learning models

Our study explores signatures for Crohn's disease (CD) and Ulcerative Colitis (UC) reflecting the interplay between the intestinal microbiota, systemic inflammation, and plasma bile acid homeostasis. For this, 1,257 individuals scheduled for colonoscopy were included and completed a comprehensive questionnaire. Individuals with IBD ('CD' n=64 and 'UC' n= 55), were age- and gender-matched to controls without findings during colonoscopy. Taxonomic profiles of the fecal microbiota and plasma profiles of inflammatory proteins and bile acids were used to build disease classifiers. Omics integration identified associations across datasets. B. hydrogenotrophica was associated with CD and C. eutactus, C. sp. CAG 167, B. cellulosilyticus, C. mitsuokai with controls. Ten inflammation markers were increased in CD, and eleven bile acids and derivatives were decreased in CD, while 7a-Hydroxy-3-oxo-4-cholestenoate (7-HOCA) and chenodeoxycholic acid (CDCA) were increased compared to controls. In UC, commensals such as F. prausnitzii and A. muciniphila were depleted. CCL11, IL-17A, and TNF were increased in UC and associated to gut microbial changes. Correlations between taxa and bile acids were all positive. For both CD and UC, taxonomic differences were primarily characterized by a reduction in commensal gut microbes which exhibited positive correlations with secondary bile acids and negative correlations with inflammation markers.