Geometric deep learning enables high-fidelity network imputation for HIV transmission modeling
Communications Medicine 2026doi:10.1038/s43856-026-01640-4
Infectious disease transmission in social and spatial context.
Individual risk factors explain only part of an epidemic. Transmission concentrates where people are connected, directly or through the places they share. With a community cohort in New Delhi, we assembled a detailed social and spatial network of 2,512 participants and the 181 venues they frequent.
These data showed that both direct and indirect ties drive the spread of HIV and hepatitis C (eLife, 2021), and phylodynamic analysis traced how hepatitis C lineages moved through the same population over space and time (Hepatology, 2021). Graph neural networks then identified seven overlapping communities, with a single venue overlapping six of them (Science Advances, 2022): a leverage point that shows where one integrated service site could reach the most people.
Current work imputes the ties that sampling misses and evaluates spatially targeted interventions under realistic coverage, so prevention can be placed by network structure rather than spread evenly across an uneven epidemic.
Communications Medicine 2026doi:10.1038/s43856-026-01640-4
Nature Communications 2026doi:10.1038/s41467-026-69874-3
Journal of Viral Hepatitis 2024doi:10.1111/jvh.13960
Journal of the International AIDS Society 2024doi:10.1002/jia2.26241
Addiction 2024doi:10.1111/add.16379
The Lancet Gastroenterology & Hepatology 2023doi:10.1016/S2468-1253(23)00018-3
Science Advances 2022doi:10.1126/sciadv.abf0158
Open Forum Infectious Diseases 2022doi:10.1093/ofid/ofac481
Scientific Reports 2021doi:10.1038/s41598-021-96779-6
eLife 2021doi:10.7554/eLife.69174
Hepatology 2021doi:10.1002/hep.31912
Clinical Infectious Diseases 2021doi:10.1093/cid/ciaa1313
The Journal of Infectious Diseases 2020doi:10.1093/infdis/jiaa006