Abstract Background Early diagnosis and risk prediction are critical to identify cases of inflammatory bowel disease (IBD) and avoid complications, such as fibrosis and tissue damage. Multiomics is key, since IBD is a complex, dynamic disease caused by genetics, immune elements, and environmental factors. The aim of this study is to integrate multiomic data and uncover mechanistic factors behind IBD. Methods Multiomic analysis was carried out using UK biobank under three conditions, all of those with IBD (n = 9,131), those diagnosed at assessment (n = 5,923) and those diagnosed after assessment (n = 3,209). Variables were screened using logistic regression of 58 clinical blood biomarkers, genetic variants, 250 blood metabolites, 2,910 blood proteins and 30 sociodemographic factors such as clinical measurements, dietary reporting, alcohol-use and smoking. Then, using matched nested case-control sampling with elastic net, random forest, multilayer perceptron, XGBoost and a stacking ensemble. Important features were collected based on feature importance in the models and combined to make a multiomic dataset that was analysed with the same algorithms. Results The most informative data type in terms of AUC, was the proteomics using the elastic net algorithm (AUC=0.762) in the diagnostic cohort. The prognostic cohort displayed reduced performance, with the multiomic data analysed with random forests most informative (AUC=0.627). Important features for diagnosis were, proteins PTGDS and GZMB, lymphocyte percentage of blood, cholesteryl esters in small HDL of metabolites and preference for white bread. In the prognostic cohort, they were proteins AGR2 and MMP12, blood cholesterol, triglycerides in small LDL of metabolites and smoking. The genetic variants most predictive were variants in FCGR2A, IFNG, and in intergenic regions of GATD3/ICOSLG, HNF4A/TTPAL and OR2H1/GABBR1. These are enriched for Autoimmune antibody positivity (p = 6.64e-3), regulation of peptide hormone secretion (p = 1.72e-2) and signalling receptor activity (p = 2.72e-2). Similarly, the union of proteins identified in models for diagnosis, were enriched for positive regulation of immune system process (p = 1.96e-8) and signalling receptor binding (p = 1.89e-6). Conclusion This study shows the benefit of combining omics data when modelling disease diagnosis and risk. Additionally, the genetic and proteomic markers are enriched for immunological characteristics, as expected in an inflammatory condition. These results provide insights into the etiology of IBD and diagnostic biomarkers. References: Cantoro, L. et al. The Earlier You Find, the Better You Treat: Red Flags for Early Diagnosis of Inflammatory Bowel. Disease. Diagnostics (Basel). 2023;13(20):3183. doi:10.3390/diagnostics13203183 Bycroft, C. et al. The UK Biobank resource with deep phenotyping and genomic data. Nature. 2018;562(7726):203-209. doi:10.1038/s41586-018-0579-z Preto AJ. et al. Multi-omics data integration identifies novel biomarkers and patient subgroups in inflammatory bowel disease. J Crohns Colitis. 2025;19(1):jjae197. doi:10.1093/ecco-jcc/jjae197 Conflict of interest: Dr. Russ, Dominic: No conflict of interest Mondal, Sudip: No conflict of interest Kollampallath, Swarnima: No conflict of interest Alzarooni, Abdulrahman: No conflict of interest Gkoutos, Georgios: No conflict of interest Acharjee, Animesh: No conflicts
Russ et al. (Thu,) studied this question.