Insights in HIV-TB co-infection with TDA Mapper and machine learning

Tuberculosis (TB) remains one of the leading causes of mortality among human immunodeficiency virus (HIV)—infected individuals. Existing models frequently fail to capture the complex dynamic relationship between TB and HIV, making it challenging to implement targeted public health policies and allocate resources efficiently. In this work, we explore the complex interactions between TB and HIV co-infection using topological data analysis (TDA). Using a dataset of 657 HIV-positive patients initiating TB treatment, the study employs TDA and machine learning models. The TDA Mapper graph constructed reveals distinct clusters of individuals that align closely with their antiretroviral therapy (ART) treatment status and history of TB infection. The Mapper graph information is used as additional feature for machine learning models. Logistic regression and linear support vector machine (SVM) models demonstrate the highest predictive accuracy. Findings underscore the necessity for early ART initiation to improve patient outcomes and offer valuable insights into improving treatment protocols or enhancing resource allocation for TB-HIV management, particularly in high-burden areas like South Africa.