Bayesian Machine Learning Techniques for revealing complex interactions among genetic and clinical factors in association with extra-intestinal Manifestations in IBD patients.

The objective of the study is to assess the predictive performance of three different techniques as classifiers for extra-intestinal manifestations in 152 patients with Crohn's disease. Naïve Bayes, Bayesian Additive Regression Trees and Bayesian Networks implemented using a Greedy Thick Thinning algorithm for learning dependencies among variables and EM algorithm for learning conditional probabilities associated to each variable are taken into account. Three sets of variables were considered: (i) disease characteristics: presentation, behavior and location (ii) risk factors: age, gender, smoke and familiarity and (iii) genetic polymorphisms of the NOD2, CD14, TNFA, IL12B, and IL1RN genes, whose involvement in Crohn's disease is known or suspected. Extra-intestinal manifestations occurred in 75 patients. Bayesian Networks achieved accuracy of 82% when considering only clinical factors and 89% when considering also genetic information, outperforming the other techniques. CD14 has a small predicting capability. Adding TNFA, IL12B to the 3020insC NOD2 variant improved the accuracy.