Developing Inference Model to Diagnosis of Primary Immunodeficiency Diseases in Protégé

Abstract

Primary immunodeficiency diseases (PIDs) are a genetically  heterogeneous group disorders that affect distinct components of both humoral and cellular arms of the immune system (1,2). Overlapping signs and symptoms of these diseases is a challenge for diagnosis and treatment (3,4). Awareness of the  symptoms and considering   the   possibility   of   PID   in   differential diagnosis help to rapid recognition and more appropriate treatment   (2,5).   Timely   recognition   and   treatment reduced mortality and increased lifespan and quality of life of the patients (6). Memorization of all effective criteria to diagnosis is difficult, so developing a computerized program based on diagnosis criteria, improves significantly the quality of care (7,8).To develop the inference model to the diagnosis of PIDs, ontology has been used in this study. The study focused on eight common diseases of PIDs include Common Variable Immune Deficiency (CVID), X- Linked Agammaglobulinemia (Bruton’s) (XLA), Selective IgA Deficiency (SIgA), CD40L deficiency, UNG deficiency, Isolated immunoglobulin (Ig) G Subclass deficiency, Specific antibody deficiency (SAD) with normal Ig concentrations and normal numbers of B cells, Transient Hypogammaglobulinemia of infancy (THI) with normal numbers of B cells. Based on clinical guidelines  and   medical   literature   in   PID   (9),   we designed a checklist to extract and classified most important signs and symptoms, family history, and laboratory data for eight main type of primary antibody deficiencies   (PADs).   To   evaluate   the   quality   of checklist, data for 100 cases in a different type of PADs were tested. Using frame-based ontology modeling to create the inference model and "Noy and McGuinness" method to develop the inference model. "Noy and McGuinness" method includes seven stages (10). Below we describe each stage of the method: