Comorbidity Disease Identification of Diabetes Insipidus Using Graph Network Algorithms

Diabetes insipidus (DI) is a one of chronic disease and with its comorbid relationship causes a serious health problem. Therefore, this study aimed to explore the effect of the known and potential comorbidity disease with DI for a better prevention and treatment strategy, moreover, compare the performance of community detection algorithms for comorbidity detection of DI. The data was collected from pubtator3 downloaded through this link: https://www.ncbi.nlm.nih'2oy/research/pubtator3/ using this following keyword “comorbid diabetes insipidus”. Then, preprocessing steps such as remove duplicate, format the data, remove N/A value from disease information, cleaning the dataset was performed. Moreover, disease ontology identifier (DOID) labelling was conducted. Furthermore, computation or modelling for network formation, community detection and comorbidity discovery were performed. All the analysis was performed using python, R and SPARQL. From 1,074 initial articles on “comorbid diabetes insipidus” 820 unique articles were identified, leading to 3,042 unique disease names mapped to 1,108 DOIDs, refined to 578 unique DOIDs. A network with 550 nodes and 2,425 edges was formed, using a 0.5 threshold. The Louvain algorithm, favored for community detection, identified 16 key comorbidities like vascular, autoimmune, and lung diseases. The consensus among centrality algorithms highlighted these comorbidity associations, with the Louvain algorithm most effective in detecting significant communities.