DOP55 High diagnostic accuracy of 4-probe methylation biomarkers in paediatric Inflammatory Bowel Disease by epigenome-wide analysis

Abstract

Abstract Background The application of -omics technology offers important opportunities for biomarker discovery to personalise management of patients with inflammatory bowel disease (IBD). In previous work, we reported a characteristic profile of genome-wide DNA methylation in peripheral blood leucocytes from children with paediatric IBD (pIBD) at diagnosis defining the IBD methylome. This characteristic pattern of genome-wide alterations was replicated in inception cohorts of adult patients in UK and Scandinavia. Methods Whole blood DNA methylation profiling was performed using the Illumina EPIC array on 86 pIBD patients and 30 non-IBD controls. Patients had a median age of 12 y and were prospectively recruited from gastroenterology clinics in Oxford and Cambridge, UK. In modelling, we utilised the paediatric BISCUIT and PICTS study cohorts from Scotland as our training data. Publicly available data from the RISK paediatric CD cohort from North America was accessed (GSE11261) to further assess accuracy of the model. The model was then subjected to further testing in an Oxford-based paediatric coeliac cohort and adult cohorts with IBD, and rheumatoid arthritis (RA)(Table 1). Results Genome-wide methylation changes in the Oxford/Cambridge cohort were highly consistent with the index BISCUIT and PICTS cohorts. Four single methylation sites were selected to make up a diagnostic model involving the genes, RPS6KA2, VMP1, CF1 and ARHGEF3 (Figure 1) in the index cohort and validated in the Oxford/Cambridge cohort. Following receiver operating characteristic (ROC) area under the curve (AUC) analyses the model demonstrated an AUC of 0.912 (95% CI: 0.86-0.96) (Table 1). To further validate our model, we compared pIBD who were CRP-positive (>5 mg/l) and CRP-negative (<5 mg/l) at presentation against non-IBD children. In the CRP-positive group, we found an AUC of 0.99 (95% CI: 0.99-1). Within the CRP-negative group an AUC of 0.90 was observed against controls (95% CI: 0.83-0.96). The model was further validated using methylation data at the baseline timepoint in the RISK cohort, with an AUC of 0.93 (95% CI: 0.90-0.96). In further analyses, we demonstrated specificity for IBD compared with paediatric coeliac disease; and accuracy higher in childhood-onset AUC than adult-onset disease AUC; the model is not accurate in diagnosis of RA. Conclusion We confirm a characteristic pattern of DNA methylation changes in childhood-onset IBD; and derive and validate a 4-probe model for diagnosis with high accuracy in both Europe and North America. The model is specific for pIBD compared with symptomatic children with no demonstrable pathology; and children with coeliac disease; and may provide an alternative to current markers in blood and stool, including calprotectin.