A Machine Learning-Based Diagnostic Model for Crohn's Disease and Ulcerative Colitis Utilizing Fecal Microbiome...

Area of Science:

• Microbiome research
• Machine learning applications in medicine
• Gastroenterology

Background:

• Fecal microbiome analysis shows potential for diagnosing inflammatory bowel disease (IBD).
• Machine learning (ML) techniques can be applied to microbiome data for disease prediction.
• Distinguishing between Crohn's disease (CD), ulcerative colitis (UC), and healthy controls (HCs) is crucial for effective IBD management.

Purpose of the Study:

• To develop and validate a machine learning model for differentiating IBD subtypes and healthy controls using fecal microbiome data.
• To assess the diagnostic performance of the sparse partial least squares discriminant analysis (sPLS-DA) model.

Main Methods:

• 16S rRNA gene sequencing was performed on fecal samples from CD (n=671), UC (n=114), and HC (n=1462) cohorts.
• A streamlined bioinformatics pipeline (HmmUFOTU) was used for data processing, retaining 1517 phylotypes and 1846 samples.
• Sparse partial least squares discriminant analysis (sPLS-DA) was employed to build binary prediction models after downsampling and data splitting.

Main Results:

• The sPLS-DA model achieved high accuracy in differentiating IBD from HC (mean accuracy=0.950, AUC=0.992).
• The model also demonstrated high accuracy in distinguishing CD from UC (mean accuracy=0.945, AUC=0.988).
• These results were validated on a separate test set derived from multicenter cohorts.

Conclusions:

• Machine learning models, specifically sPLS-DA, based on fecal microbiome analysis hold significant diagnostic potential for IBD.
• The developed model can effectively differentiate between IBD and healthy individuals.
• The model also shows strong capability in distinguishing between Crohn's disease and ulcerative colitis.