Linking Gene Expression to Clinical Outcomes in Pediatric Crohn's Disease Using Machine Learning

Abstract

IntroductionPediatric Crohns disease (CD) is the fastest growing age group and is characterized by frequent disease complications. We sought to analyze both ileal and colonic gene expression in a cohort of pediatric CD patients and apply machine learning-based models to predict risk of developing future complications. MethodsRNA-seq was generated from matched ileal and colonic biopsies from formalin-fixed, paraffin-embedded (FFPE) tissue obtained from patients with non-stricturing/non-penetrating, treatment-naive CD and from controls. Clinical outcomes including development of strictures or fistulas, progression to surgery, and remission were analyzed first using differential expression. Machine learning models were then developed for each outcome, combining gene expression and clinical factors. Models were assessed using area under the receiver operating characteristic curve (AUROC). Results56 patients with CD and 46 controls were included. Differential expression analysis revealed a distinct colonic transcriptome for patients who developed strictures, with downregulation of pathways related to inflammation and extra-cellular matrix production. In contrast, there were few differentially expressed genes for other outcomes and for ileal tissue. Despite this, machine learning-based models were able to incorporate colonic gene expression and clinical characteristics to predict outcomes with high accuracy. Models showed an AUROC of 0.84 for strictures, 0.83 for remission, and 0.75 for surgery. Certain genes with potential prognostic importance for strictures (REG1A, MMP3, and DUOX2) were not identified in single gene differential analysis but were found to have strong contributions to predictive models. ConclusionsOur findings in FFPE tissue support the importance of colonic gene expression and the potential for machine learning-based models in predicting outcomes for pediatric CD.