Delicia Wong, Western University

Delicia Wong, William Wang, Art FY Poon — Western University, London, ON, Canada

Discovering & Evolving

Phylodynamic methods estimate key epidemiological parameters, including the basic reproduction number (R₀), from viral sequence data by linking phylogenetic tree structure to transmission dynamics. These approaches generally assume sampled sequences are related by a single phylogenetic tree. However, superinfection allows divergent viral lineages to co-infect the same host and undergo recombination—the exchange of genetic material between lineages widespread among RNA viruses. Recombination generates discordant evolutionary histories across the genome that cannot be represented by a single tree, potentially biasing phylodynamic inference. We quantified the impact of superinfection and recombination on R₀ estimates using a multi-level simulation framework. Transmission trees and nested within-host phylogenies were simulated under a susceptible–infected–removed (SIR) model. Phylodynamic inference used the birth–death SIR (BDSIR) model in BEAST2. Simulations without recombination established baseline accuracy. Recombination and superinfection were introduced by generating ancestral recombination graphs resolved into local phylogenies between breakpoints. Sequence alignments (100 tips) were simulated using pyvolve and analyzed with BDSIR. Inferred R₀ values were compared to known simulation parameters (R₀ = 1.96 for HIV-1; 3.38 for SARS-CoV-2). Baseline simulations accurately recovered R₀ without recombination (RMSE = 0.06 for HIV-1; 0.12 for SARS-CoV-2). With increasing recombination, reconstructed trees became increasingly star-like, with shorter internal and longer terminal branches. Correspondingly, R₀ estimates were significantly biased upward by 0.012 (95% CI 0.007–0.016) per 10 breakpoints for HIV-1 and 0.041 (0.030–0.053) for SARS-CoV-2. These results demonstrate that recombination substantially inflates phylodynamic estimates of transmission potential, highlighting limitations of standard tree-based methods for recombinogenic viruses.