Inflammatory bowel diseases (IBD), Crohn’s disease (CD), Celiac’s disease (CeD) and ulcerative colitis (UC) are chronic relapsing remitting inflammatory conditions affecting the gastrointestinal tract, primarily the small intestine and colon . CD is most frequently diagnosed in patients in their 20s and UC in their 30s; however, the diagnosis can be made at any age . IBD diagnosis is often straightforward, as disease can be seen by endoscopy or imaging modalities. However, diagnosis can be difficult as patients may experience symptoms consistent with IBD but ultimately have other diagnoses including functional gastrointestinal disorders such as irritable bowel syndrome (IBS) [3–6]. Patients with IBS can have symptoms very similar to those with IBD. IBD can be limited to difficult to evaluate areas of the GI tract such as isolated small bowel disease. Also, within IBD, differentiating between CD and UC can be difficult, especially within patients with severe inflammatory activity, often termed indeterminate colitis. When the clinical presentation is severe and an operation including colectomy is indicated, differentiating CD and UC is imperative, as ileal pouch-anal anastomosis (IPAA) is generally contraindicated in CD due to high morbidity .
Developing biomarkers that can be easily obtained and allow for the correct diagnosis early into evaluation can avoid costly interventions that expose patients to multiple unnecessary procedures. Blood markers for both IBD and IBS have been sought for decades. For IBD, perinuclear antineutrophil cytoplasmic antibody (p-ANCA) and anti-Saccharomyces cerevisiae antibody (ASCA) have been reported to be markers for UC and CD, respectively. However, p-ANCA is also detected in 10–40% of patients with CD and ASCA is detected in 6–14% of patients with UC . Other markers increased in subjects with CD include antibodies to (a) Escherichia coli outer membrane porin C (Omp-C), (b) protein from Pseudomonas fluorescens and (c) flagellin c-BIR1 (anti-CBIR1) , but these markers remain insensitive. In patients with indeterminate colitis, those with one or more positive antibodies, including ANCA, ASCA, I2 (antibody to Pseudomonas fluorescens), and Omp-C, have significantly higher post-operative complications . Other inflammatory biomarkers such as C-reactive protein, fecal calprotectin, and fecal lactoferrin differentiate IBD from other gastrointestinal disorders such as IBS , but tests do not differentiate among various types of inflammatory colitides .
Additional biomarker candidates include DNA variants, differences in RNA transcript abundances, including mRNAs, microRNAs, long intergenic non-coding RNAs, proteins, or metabolites. A general view is that different profiles of biomarkers could provide useful information to guide clinical decision-making; from diagnosis to choice of optimal therapies and in some cases these biomarker profiles are being implemented in clinical practice [3, 12–24]. Searches for optimal biomarker profiles can be achieved using clustering methods e.g., heirarchical clustering, K-means clustering, which depend upon the general ability to find common features across a sample population or forms of linear discriminate analysis, which depend upon the ability to find linear combinations of features that have the ability to separate two or more classes. The former method is a common method to analyze large numbers of features, such as microarray data whereas the latter is a more common method for analysis of smaller numbers of features. Both methods are suitable for further analyses using machine learning methods such as support vector machines, logistic regression, principal components analysis or prediction analysis for microarrays. Using a form of linear discriminant analysis, we have attempted to employ mRNA transcript profiles to distinguish between subjects with multiple sclerosis and other comparator groups [25, 26]. Our results clearly demonstrate that mRNA transcript profiling has the capacity to distinguish between MS, even early in the disease process, and homogeneous comparator groups, such as healthy subjects (CTRL), or subjects with clinically related diseases such as neuromyelitis optica or transverse myelitis. Thus, these binary comparisons can produce a test of exclusion of multiple sclerosis. Here, we applied this approach to IBD and IBS. Our results demonstrate that distinct mRNA profiles accurately discriminate IBD from CTRL, IBS from CTRL, IBD from IBS, and CD from UC with high degrees of sensitivity and specificity. We propose these approaches may provide useful guides for clinical decision-making.