A Novel Tool for Celiac Disease Classification Based on Clinical, Immunological and Genetic Markers

Celiac Disease (CD) is an autoimmune disease characterized by the demolition of the intestinal parietal cells by the immune system triggered by the ingestion of gluten. Although it is a relatively common condition, its etiology is mostly unknown, and it can only be treated with the complete exclusion of gluten from the diet. Currently, several subtypes of CD have been identified, based on its clinical presentation, defined as Major, Minor, Silent, Latent and Potential. Some studies have shown that the "Major" type is associated with an increased probability of complications. In fact, if not properly treated, CD can lead to severe complications, therefore it is important to make a timely diagnosis and identify the proper subtype. However, symptoms are often nonspecific, mild or absent, and the characterization of the disease is very complex since it depends on a combination of numerous clinical, demographic, genetic and immunologic factors. In this short study, we present a tool that can be used to automatically classify celiac patients based on their clinical, immunological, histologic, and genetic characteristics. This tool features an interactive graphical user interface and is based on a supervised machine learning model, i.e. a Stacking Classifier that integrates three models: Gradient Boosting, Random Forest, and Decision Trees. The classifier is trained on a large multicentric Italian cohort, containing almost 2,500 patients. Supporting the identification of the right CD subtype can be an important step towards the prevention of serious complications through early treatment and screening strategies.