Immunological Profiling of Paediatric Inflammatory Bowel Disease Using Unsupervised Machine Learning.

Area of Science:

• Immunology
• Gastroenterology
• Computational Biology

Background:

• Current inflammatory bowel disease (IBD) classification relies on clinical phenotypes, overlooking molecular disease mechanisms.
• Stratifying IBD patients based on molecular profiles is crucial for personalized treatment approaches.

Purpose of the Study:

• To stratify a treatment-naïve pediatric IBD cohort using innate immunity pathway profiling.
• To apply unsupervised machine learning (UML) for identifying distinct patient subgroups based on immune responses.

Main Methods:

• Assessed innate immune responses in pediatric IBD patients and controls by stimulating peripheral blood mononuclear cells (PBMCs).
• Measured cytokine responses (IL-10, IL-1β, IL-6, TNF-α) following stimulation of NOD2, TLR1-2, and TLR4 pathways.
• Utilized hierarchical clustering on cytokine data for patient stratification.

Main Results:

• Pediatric IBD patients exhibited significantly reduced combined immune responses compared to controls.
• Hypofunctional toll-like receptor (TLR) responses were identified as a primary driver of reduced immune function.
• Hierarchical clustering revealed 3 distinct patient clusters and an unclustered group, independent of clinical disease phenotype.

Conclusions:

• While hierarchical clustering did not yield clinically actionable outcomes, it supports UML as a valid approach for patient stratification in IBD.
• The study underscores the prevalence of hypo-inflammatory innate immune responses in the pathogenesis of IBD.
• Further research into immune profiling could refine IBD patient stratification and therapeutic strategies.