Meta analysis of the relationship between celiac disease and immune response

Celiac disease is an autoimmune disease triggered by gluten found in grains such as wheat, barley and rye, characterized by chronic intestinal inflammation and villus atrophy in genetically susceptible individuals. Although the incidence is affected by geographical conditions and ethnic structure of societies, it is between 1% and 2% worldwide. Patients usually have typical symptoms such as abdominal pain, diarrhea and vomiting, as well as atypical symptoms such as retardation of growth, anemia, fatigue, irritability, and dermatological problems. Clinical applications are highly challenging because of the high variability in patient profile, and non-standardized serological tests. The disease is closely related to the immune system. The immune system's response to gluten is first determined by detecting the relevant autoantibodies in the blood. The definitive diagnosis of the disease reached by investigating the degree of histological changes in the villus as a result of the destruction of the small intestine by the immune system mechanisms triggered META ANALYSIS OF THE RELATIONSHIP BETWEEN CELIAC DISEASE AND IMMUN RESPONSE xiv by gluten. Serological tests are used in the first stage of diagnosis. If the serological tests are positive, the patient is directed to the intestinal biopsy. Therefore, diagnostic accuracy parameters such as sensitivity, specificity and predictive values of serological tests are very important in the biopsy decision. However, these values are not given in most clinical studies. These missing parameters complicate the standardization of serological tests, and diagnosis. Although there are traditional meta-analysis studies examining the diagnostic accuracy of celiac serological tests in the literature, no study has been found, evaluating the relationship between celiac disease and the immune system by using machine learning algorithms. The aim of this thesis is to develop a different meta-analysis approach that estimates the diagnostic accuracy parameters of serological tests used in the measurement of autoantibodies, one of the immune indicators of the disease, with machine learning algorithms. For this purpose, four different machine learning algorithms: decision trees, random forest, gradient boosting and naive Bayes classification were applied to the datasets of anti-gliadin antibody (AGA), anti-endomysium antibody and anti-tissue transglutaminase antibody. It was found that the machine learning algorithm that most successfully predicted the sensitivity parameters of the immunological tests used in the diagnosis of celiac was the decision trees algorithm with an accuracy of 88.7%. It is thought that the diagnostic accuracy parameters estimated by this algorithm will contribute to the standardization of serological tests and can tolerate missing data in epidemiological studies and meta-analyses.