Abstract:ABSTRACT Objective We have previously described a prognostic transcriptional signature in CD8 T cells that separates inflammatory bowel disease (IBD) patients into 2 phenotypically-distinct subgroups, which we termed IBD1 and IBD2. Here we sought to develop a blood-based prognostic test that could identify these patient subgroups without the need for cell separation, and thus be suitable for routine clinical use in Crohn’s disease (CD) and ulcerative colitis (UC). Design Patients with active IBD were recruited before treatment. Transcriptomic analyses were performed on purified CD8 T cells and/or whole blood. Detailed phenotype information was collected prospectively. IBD1/IBD2 patient subgroups were identified by consensus clustering of CD8 T cell transcriptomes. In a training cohort, statistical (machine) learning was used to identify groups of genes (“classifiers”) whose differential expression in whole blood re-created the IBD1/IBD2 patient subgroups. Genes from the best classifiers were qPCR-optimised, and further statistical learning was applied to the qPCR dataset to identify the optimal classifier, which was locked-down for further testing. Independent validation was sought in separate cohorts of CD (n=66) and UC patients (n=57). Results In both independent validation cohorts, a 17-gene qPCR-based classifier stratified patients into two distinct subgroups. Irrespective of the underlying diagnosis, IBDhi patients (analogous to the poor prognosis IBD1 subgroup) experienced significantly more aggressive disease than IBDlo patients (analogous to IBD2), with earlier need for treatment escalation and more escalations over time. Conclusion This is the first validated prognostic biomarker that can predict prognosis in newly-diagnosed IBD patients, and represents a step towards personalised therapy. What is already known about this subject? The course of CD and UC varies considerably between patients, but reliable prognostic markers are not available in clinical practice. This hinders disease management because treatment approaches that would be optimal for patients with indolent disease – characterised by infrequent flare-ups that can be readily controlled by first-line therapy – will inevitably undertreat those with progressive disease. Conversely, strategies that would appropriately control frequently-relapsing, progressive disease will expose patients with more quiescent disease to the risks and side effects of unnecessary treatment. We have previously described a CD8 T cell gene expression signature that corresponds to differences in T cell exhaustion, is detectable during active untreated disease (including at diagnosis), and predicts disease course in both UC and CD. However, the need for cell separation and microarray-based gene expression analysis would make this difficult to translate to clinical practice. What are the new findings? We have developed, optimised, and independently validated a whole blood qPCR-based classifier – designed to identify the IBD1 and IBD2 patient subgroups – that can reliably predict prognosis in patients with CD or UC from diagnosis without the need for cell separation. We also present a detailed phenotypic update on the disease course experienced by patients in either the IBD1/IBDhi or IBD2/IBDlo subgroups, incorporating both expanded patient cohorts and substantially longer follow-up. This affords new insights into the spectrum of therapies that are differentially required in these patient subgroups, and reinforces their association with disease prognosis. How might it impact on clinical practice in the foreseeable future? The qPCR-based classifier has performance characteristics that compare favourably to prognostic biomarkers currently in use in oncology, and should be sufficient to guide therapy from diagnosis in patients with CD or UC. This represents an important step towards personalised therapy in inflammatory bowel disease.

P455 Deploying machine learning algorithms to stratify patients with inflammatory bowel disease using routinely collected electronic health records

P070 Machine learning approaches to identify IBD biomarkers from longitudinal microbiome data

Large-scale clustering of longitudinal faecal calprotectin and C-reactive protein profiles in inflammatory bowel disease

Predicting inpatient mortality in patients with inflammatory bowel disease: A machine learning approach

Artificial Intelligence for Inflammatory Bowel Diseases (IBD); Accurately Predicting Adverse Outcomes Using Machine Learning

Classification of Paediatric Inflammatory Bowel Disease using Machine Learning

Supervised machine learning classifies inflammatory bowel disease patients by subtype using whole exome sequencing data

P160 Integrative transcriptional analysis to identify predictive biomarkers for IBD severity and subtyping

P488 Development and validation of a risk prediction tool for inflammatory bowel disease (IBD) diagnosis in patients presenting in primary care with abdominal symptoms, a UK based study

A blood-based prognostic biomarker in inflammatory bowel disease

Gene Identification in Inflammatory Bowel Disease via a Machine Learning Approach

Machine Learning Prediction Model for Inflammatory Bowel Disease Based on Laboratory Markers. Working Model in a Discovery Cohort Study

Performance of Machine Learning Algorithms for Predicting Disease Activity in Inflammatory Bowel Disease

Can molecular stratification improve the treatment of inflammatory bowel disease?

Predictive, preventive and personalised approach as a conceptual and technological innovation in primary and secondary care of inflammatory bowel disease benefiting affected individuals and populations

P503 Differentiation of patients with IBD at increased risk of CRC using UR-CARE: A Single Centre Experience

Simplified Machine Learning Models Can Accurately Identify High-Need High-Cost Patients With Inflammatory Bowel Disease

The predictive ability of inflammatory biomarkers in patients with newly diagnosed Inflammatory Bowel Disease [IBD]: Analyses of preliminary pre‐ and post‐therapy data

DOP31 Serum protein markers for early and differential IBD diagnosis validated by machine learning approaches

Evaluating Inflammatory Bowel Disease-Related Quality of Life Using an Interpretable Machine Learning Approach: A Multicenter Study in China

IIS Registry Grant Modelling Inflammatory Bowel Diseases trajectories combining dynamic, multifactorial, Artificial Intelligence-based approaches