Decoding the Molecular Landscape of Inflammatory Bowel Disease: A Mendelian Randomization and Machine Learning...

Area of Science:

• Genetics
• Metabolomics
• Immunology

Background:

• Inflammatory bowel disease (IBD) presents a growing global health challenge.
• IBD pathogenesis involves complex interactions of genetic, environmental, and metabolic factors.
• Understanding plasma protein associations is crucial for IBD research.

Purpose of the Study:

• To investigate causal relationships between plasma proteins and IBD risk.
• To identify novel biomarkers and therapeutic targets for IBD.
• To leverage Mendelian randomization and machine learning for etiological insights.

Main Methods:

• Two-phase Mendelian randomization (MR) using genetic variants as instrumental variables.
• Summary-data-based MR and machine learning models applied to large-scale genetic and proteomic data.
• Validation in independent cohorts (Pan-UKB, IIBDGC) and SMR-Phenome-Wide Association Studies (SMR-PheWAS).

Main Results:

• Several plasma proteins were significantly associated with IBD risk, confirmed across cohorts.
• Machine learning identified key genes (e.g., FCGR3A, MAPKAPK2) for Crohn's disease and ulcerative colitis prediction.
• Potential therapeutic targets include GCKR, MSP, and FCG2A, with MSP and GCKR showing causal links to other conditions.

Conclusions:

• The study provides novel insights into IBD pathogenesis using advanced statistical methods.
• Identified proteins and genes offer potential for precision medicine in IBD diagnosis and treatment.
• Findings highlight the need for validation in diverse populations to broaden applicability.