TRACE-5.0: Dynamic Transmission Surveillance, or How HIV-TRACE Learned to Stop Worrying and Love the Molecular Clock

Inferring transmission networks from viral sequences routinely collected during antiretroviral drug-resistance screening is a pillar of modern human immunodeficiency virus type 1 (HIV-1) surveillance. Public health departments use these networks to guide prevention and public health services. Standard approaches cluster sequence pairs below a fixed genetic-distance threshold (typically 0.5% or 1.5%) and use recent cluster growth to prioritize outbreaks. However, ignoring elapsed time between sampling dates and HIV-1 evolutionary dynamics causes static distance thresholds to sever delayed transmission links, merge unlinked lineages into sprawling clusters, and mistake sequencing or sample-labeling errors for active outbreaks. Here we present TRACE-5.0, a framework that retains the speed and simplicity of pairwise distance clustering while incorporating sampling dates into explicit molecular-clock and epidemiological models. At the pairwise scale, DANNO evaluates genetic divergence against elapsed sampling time to certify transmission links and reject clock violations. At the cluster and population scales, STEVE partitions inferred transmission networks by local evolutionary rate to prioritize rapidly expanding outbreaks, while CHIN estimates unobserved transmitting population sizes from network collision rates. Across four cohorts comprising 10,259 individuals, TRACE-5.0 eliminates specimen-contamination clusters and adapts to low-diversity founder epidemics. In temporal-holdout validation in Washington, DC, DANNO nearly doubles resource-normalized yield from 0.400 to 0.796 linked follow-up cases per prioritized patient while reducing patient investigations by 61.4%. Concurrently, STEVE achieves 83.3% cluster-level precision while cutting non-growing prioritized clusters by 92.9%. TRACE-5.0 runs locally in any modern web browser without installation (https://github.com/veg/trace-5.0), screening nationwide registries in seconds.