Umeå universitets logga

Background & Aims: Biomarkers are needed to identify individuals at elevated risk of inflammatory bowel disease. This study aimed to identify protein signatures predictive of inflammatory bowel disease. Methods: Using large population-based cohorts (n ≥180,000), blood samples were obtained from individuals who later in life were diagnosed with inflammatory bowel disease and compared with age and sex-matched controls, free from inflammatory bowel disease during follow-up. A total of 178 proteins were measured on Olink platforms. We used machine-learning methods to identify protein signatures of preclinical disease in the discovery cohort (n = 312). Their performance was validated in an external preclinical cohort (n = 222) and assessed in an inception cohort (n = 144) and a preclinical twin cohort (n = 102).

Results: In the discovery cohort, a signature of 29 proteins differentiated preclinical Crohn's disease (CD) cases from controls, with an area under the curve (AUC) of 0.85. Its performance was confirmed in the preclinical validation (AUC = 0.87) and the inception cohort (AUC = 1.0). In preclinical samples, downregulated (but not upregulated) proteins related to gut barrier integrity and macrophage functionality correlated with time to diagnosis of CD. The preclinical ulcerative colitis signature had a significant, albeit lower, predictive ability in the discovery (AUC = 0.77), validation (AUC = 0.67), and inception cohorts (AUC = 0.95). The preclinical signature for CD demonstrated an AUC of 0.89 when comparing twins with preclinical CD with matched external healthy twins, but its predictive ability was lower (AUC = 0.58; P = .04) when comparing them with their healthy twin siblings, that is, when accounting for genetic and shared environmental factors.

Conclusion: We identified protein signatures for predicting a future diagnosis of CD and ulcerative colitis, validated across independent cohorts. In the context of CD, the signature offers potential for early prediction.