P307 Novel proteomic analysis of 11,000 serum proteins distinguishes between Crohn’s disease and ulcerative colitis in pediatric patients

Abstract

Abstract Background While the two main inflammatory bowel diseases (IBD), Crohn’s disease (CD) and ulcerative colitis (UC), share various clinical features, they differ in clinical, endoscopic and histological features, and may have different underlying etiopathological mechanisms, resulting in different treatment options. Consequently, the development of reliable, highly accurate, non-invasive biomarkers that differentiate CD and UC in pediatric IBD are an unmet need of paramount importance in terms of medical care, surgical intervention, and prognosis. The most advanced proteomics technology, the aptamer-based SomaScan, enables measurement of 11,000 biologically relevant proteins, including very low abundant proteins, and was tested to discover serum protein biomarkers capable of differentiating between pediatric CD and UC. Methods SomaScan (SomaLogic; Boulder, CO) quantitative serum protein profiles were generated from pediatric subjects diagnosed with CD (n=56) and UC (n=25) using standard diagnostic criteria. The disease location for patients with CD was 18 L1 (ileal), 9 L2 (colonic), 27 L3 (ileo-colonic) and 2 L4 (small bowel); Inflammatory phenotype (B1) occurred in 40, stricturing (B2) in 8, penetrating (B3) in 6, and B2B3 in 2. To develop highaccuracy predictors with the lowest number of proteins for discriminating CD from UC, we used machine learning and conditional logistic regression to calculate odds ratios(OR) and 95% confidence intervals(CI) per one standard deviation increase in protein levels. Area under the curve (AUC) was calculated to determine the performance of the multi-protein model in discriminating CD cases from UC cases. Ingenuity pathway analysis was conducted to identify pathophysiological pathway differences between CD and UC. Results There were 256 proteins that were discriminating between CD and UC with fold change >1.3 and an unadjusted p-value <0.05. Of these, 38 proteins were associated with increased odds and 218 proteins were associated with decreased odds of CD diagnosis. Compared to UC, CD cases had decreased neutrophil movement and degranulation, inflammatory response, and metabolism of reactive oxygen species but enhanced apoptosis. When we developed a multi-protein model and assessed its performance to discriminate CDs cases from UC cases, a 4-protein model showed an AUC=0.97. Conclusion Using the most comprehensive proteomics platform, we identified serum proteins and developed a high accuracy multi-protein model discriminating between pediatric CD and UC. The utility of SomaScan demonstrates not only the value of these diagnostic biomarkers, but also the potential to discover immune and metabolic pathways that distinguish CD from UC.