Forecasting trajectories of HIV transmission networks with a novel phylodynamic and deep learning framework

Co-PI: Dapeng Oliver Wu

Sponsor: National Institutes of Health

Start Date: May 1, 2020

End Date: April 30, 2024

Amount: $2,809,658

Abstract

University of Florida researchers are crossing disciplinary lines to harness the power of artificial intelligence and predict where future HIV/AIDS hot spots may emerge. Professors Mattia Prosperi, of UF College of Public Health and Health Professions; Marco Salemi, of UF College of Medicine; and Dapeng Oliver Wu, of UF Herbert Wertheim College of Engineering, will use a specialized AI technique known as deep learning to study patterns of HIV transmission. Deep learning methods use artificial neural networks that learn from complex data sets. The researchers plan to identify social, demographic, and behavioral risk profiles that will enable more powerful predictions about future trends, including where transmission clusters are likely to occur.

Traditional tools to track and mitigate HIV transmission clusters involve genomic sequencing and analyses of how the virus changes within populations and across time. Researchers can reconstruct transmission chains by looking backward at the degree of relatedness between viral strains and how they spread. But the UF researchers want to create a new tool that looks to the future. They plan to create a model that will couple existing socio-demographic and phylodynamic data with artificial intelligence that can monitor HIV transmission clusters nearly in real-time. The idea is to combine existing data sets and information about transmission with artificial intelligence capable of predicting the future trajectory of localized epidemics.

Their results would ideally better inform regional and state health officials about where to allocate resources in efforts to reduce the spread of HIV and AIDS.