Long, Non-coding RNA Gene Expression Signatures to Distinguish Irritable Bowel Syndrome and Inflammatory Bowel Disease

Abstract

BACKGROUND: lncRNAs play critical roles in the regulation gene activity. This extends to genes whose protein products are critical for mounting both innate and adaptive immune responses. Our understanding of the functional role of lncRNAs in human diseases, including gastrointestinal disease, is in its infancy. No blood-based RNA biomarkers have been made commercially available to distinguish IBS from IBD or identify individual inflammatory colitides. Here we sought to develop machine learning classifiers using long, non-coding RNA (lncRNA) gene expression data from blood to distinguish irritable bowel syndrome (IBS) and inflammatory bowel disease (IBD). METHODS: Peripheral whole blood collected into PAXgene tubes was obtained from healthy control subjects (n=115), and patients diagnosed with irritable bowel syndrome (n=128), Crohn's disease (n=89), and ulcerative colitis (n=84). Patients diagnosed with celiac disease (n=39) were recruited as an additional inflammatory disease control group for classifiers capable of distinguishing IBS from Crohn's disease or ulcerative colitis. RNA sequencing was performed using a small subset of healthy control, celiac, Crohn's disease and ulcerative colitis patient samples to derive 48 highly differentially expressed candidate lncRNA gene targets. To validate sequencing findings, RT-PCR was performed on all patients recruited in the study (n=489). Gene expression datasets generated were used to train and independently validate machine learning classifiers capable of distinguishing IBS and IBD from other subjects in the study cohort. RESULTS: lncRNAs measured by RT-PCR exhibit high degrees of differential expression across healthy control, IBS, and IBD cohorts. Unlike previous studies of mRNAs, lncRNA expression differences were frequently 4-fold or greater in case/control comparisons. lncRNAs exhibit a high degree of discriminatory power and confidence of machine learning predictions with accuracy exceeding 90% for classifiers capable of discriminating irritable bowel syndrome from other inflammatory conditions and healthy controls. CONCLUSION(S): Gene expression data derived from peripheral whole blood analyzed using machine learning methods produces classifiers capable of distinguishing presence of irritable bowel syndrome and inflammatory bowel disease. Use of this information may provide clinically actionable information for healthcare providers.