Disease Similarity Measurement Implementation for Atlas of Human Infectious Diseases Using BIOSSES

Over the past 30 years, more than 30 emerging infectious diseases (EIDs) have surfaced globally, necessitating coordinated efforts at the international level to better prevent and treat these diseases. This includes utilizing knowledge resources such as the Atlas of Human Infectious Diseases (AHID). The information contained in AHID has been structured and summarized in previous research in the form of a web-based dictionary, but there is currently no visualization representing the similarities between diseases. Disease similarity analysis is important in understanding the pathogenesis of complex diseases, early prevention, diagnosis of major diseases, and even the development of new drugs. Based on the AHID data, which contains biomedical text, this study applied the Biomedical Text Semantic Sentence Similarity Estimation System (BIOSSES) method to measure disease similarity using attributes such as epidemiology, clinical findings, agent, transmission, incubation period, and diagnostic tests. This method achieved a Pearson Correlation Coefficient (PCC) value of 0.836 in previous research. However, the model produced a low text similarity score with a PCC value of 0.3630 and a Median Absolute Deviation (MAD) of 0.1158, indicating underfitting. The results of the similarity score measurement is presented in the form of a human disease network and are available on the web application alongside the AHID Dictionary.