Improving Differentiation of Crohn's Disease and Ulcerative Colitis Proteomes through Protein-Wide Association Study...

Area of Science:

• Proteomics
• Genomics
• Biomarker Discovery

Background:

• Accurate diagnosis of Crohn's disease (CD) and ulcerative colitis (UC) is critical for effective treatment of inflammatory bowel disease (IBD).
• Current diagnostic methods, including endoscopy and histology, are invasive and expensive.
• Novel, non-invasive biomarkers are needed for improved differentiation between CD and UC.

Purpose of the Study:

• To identify plasma proteomic signatures for differentiating CD from UC using a Protein-Wide Association Study (PWAS) approach.
• To evaluate the diagnostic utility of these identified protein signatures in machine learning (ML) classifiers.
• To explore the potential of novel protein biomarkers for IBD diagnosis.

Main Methods:

• Plasma protein levels of 2,920 proteins were analyzed in 1,106 participants (636 CD, 470 UC) from the SPARC cohort.
• A PWAS with Bonferroni correction identified proteins associated with CD vs. UC, controlling for age, sex, and disease severity.
• Random forest ML models were trained using PWAS-identified proteins and evaluated using ROC-AUC; feature importance was assessed via SHAP analysis.

Main Results:

• Thirteen proteins were significantly differentially abundant between CD and UC (adjusted p < 8.42E-06).
• ML models using PWAS-identified proteins achieved an average ROC-AUC of 0.73, outperforming models using the full proteome (0.62).
• Granzyme B, insulin-like peptide 5 (INSL5), and interleukin-12 subunit beta (IL-12B) were identified as key differentiating features.

Conclusions:

• PWAS-based feature selection effectively identifies diagnostic biomarkers in complex proteomic datasets.
• Novel protein biomarkers, such as INSL5, can potentially complement existing diagnostic strategies for differentiating CD and UC.
• This study highlights the potential of plasma proteomics for non-invasive IBD diagnosis and management.