Article Text
PDF
Abstract
Background Bile acid metabolism is involved in the development of various intestinal diseases, but whether the altered bile acid metabolism is disease-specific and individually heterogeneous remains unclear. We aim to uncover Crohn’s Disease (CD) -specific bile acid metabolism features and to explore the clinical application of bile acids in CD.
Methods A total of 1255 patients with various intestinal diseases and 154 healthy individuals were enrolled. The concentrations of 18 bile acids in serum were detected using liquid chromatography-tandem mass spectrometry. Based on k-means hierarchical clustering, all subjects were classified into two kinds of bile acid metabolism subtypes. CD-specific bile acid metabolic features were identified. Predictive models were developed using five machine learning methods to assess the predictive efficacy of bile acids in the diagnosis of CD.
Results A total of 90 bile acid-related parameters were included in the study. There was a significant difference in bile acid distribution between intestinal disease patients and healthy individuals (HC). Compared to healthy controls, the greatest number of differential bile acids was observed in CD and ulcerative colitis (UC) (IDDF2024-ABS-0286 Figure 1. Principal component analysis and heatmap of serum bile acids profile). Cluster analysis identified two bile acid metabolism subtypes (BM): most HC belonged to BM1, while BM2 was associated with significantly elevated inflammatory markers such as CRP and serum amyloid A (IDDF2024-ABS-0286 Figure 2. Bile acid metabolism subtyping BM and clinical characteristics). Subtype analysis based on metabolism typing revealed that the dysregulation of conjugated bile acid metabolism pathways was a specific feature in CD (IDDF2024-ABS-0286 Figure 3. BM subtype analysis of Crohn’s disease patients). Compared to CD patients of BM1, those of BM2 had higher CRP (odds ratio 1.36, P=0.004) and a higher proportion of penetrating CD (odds ratio 1.85, P=0.001) (IDDF2024-ABS-0286 Figure 4. Multiple logistic regression analysis of BM in Crohn’s disease patients). Finally, the predictive model analysis demonstrated that a random forest model integrating three bile acids could effectively diagnose penetrating CD patients with an area under the curve of 0.82 (IDDF2024-ABS-0286 Figure 5. Bile acids predictive models developed using five machine learning methods).
Principal component analysis and heatmap of serum bile acids profile.
Bile acid metabolism subtyping BM and clinical characteristics
BM subtype analysis of Crohn’s disease patients.
Multiple logistic regression analysis of BM in Crohn’s disease patients.
Bile acids predictive models developed using five machine learning methods.
Conclusions The BM subtyping provides improved discrimination among various intestinal disease patients and healthy individuals. The dysregulation of conjugated bile acid metabolism emerges as one specific metabolic feature of CD. A predictive model integrating three bile acids can effectively diagnose patients with penetrating CD.