Tree Reconstruction Errors Can Inflate Reassortment Counts in Avian Influenza

Reassortment, the exchange of entire genome segments between influenza A viruses, has driven some of the most consequential events in viral evolution, including the antigenic shifts that introduced novel genomic combinations into circulating strains. Because segmented viruses can swap whole segments, researchers typically detect reassortment by looking for discordance between phylogenetic trees built from different genomic segments, often measured as subtree prune and regraft (SPR) rearrangements between segment trees. A study by Hugo G. Castelán-Sánchez and Art F. Y. Poon asks how much of this discordance is genuine reassortment and how much is an artifact of tree reconstruction error. When a segment’s sequence data carry too little information to resolve its evolutionary history accurately, the reconstructed tree can differ from trees of other segments for reasons that have nothing to do with reassortment, producing false signals.

To quantify this problem, the researchers curated a dataset of 11,795 complete avian H5Nx influenza A virus genomes. From these, they sampled subsets of hemagglutinin (HA) sequences chosen to represent different expected levels of phylogenetic reconstruction error, ranging from randomly selected samples to the most genetically divergent subsets. The key to their approach was simulation: the team simulated molecular evolution along the trees reconstructed from each HA sample, varying rates of evolution and sequence lengths so that the simulated data mimicked the other genomic segments of the virus. Crucially, these simulations contained no reassortment at all. That design choice gives the analysis its power, because any phylogenetic discordance detected among trees reconstructed from the simulated sequences must, by definition, be caused by reconstruction error rather than by genuine segment exchange.

The results were striking. Simulations based on random samples of HA sequences yielded roughly 32% as many SPR rearrangements as the researchers obtained from the real avian influenza data. In other words, a substantial fraction of the discordance signal that would ordinarily be interpreted as reassortment can be reproduced by reconstruction error alone when sampling is not carefully controlled. Sampling design mattered enormously: when the team simulated trees from the most genetically divergent subset of HA sequences, the false positive rate dropped to about 4%. This suggests that datasets rich in genetic diversity provide enough phylogenetic signal to reconstruct segment trees accurately, whereas datasets drawn without attention to divergence can generate misleading discordance. The researchers also examined the opposite failure mode, missing real reassortment events. By deliberately modifying simulated trees with SPR rearrangements, they measured a false negative rate of 8.5%. Notably, this rate increased as the number of SPRs per tree grew, meaning that heavily reshuffled genomes were more likely to evade detection than lightly reshuffled ones.

Taken together, the findings indicate that reassortment can be substantially overestimated when inferred purely from phylogenetic discordance. The magnitude of the effect depends on both the genetic composition of the sampled sequences and the amount of information available for tree reconstruction, two factors that vary widely across influenza surveillance datasets. The authors emphasize that analyses of influenza A virus genomes for reassortment should account for phylogenetic uncertainty and for sampling design when interpreting the evolutionary history of these viruses.

Subject of Research: The effect of phylogenetic reconstruction error on genetic reassortment inference in avian influenza A virus genomes

Article Title: Assessing the impact of phylogenetic reconstruction error on genetic reassortment inference in segmented viruses

Article References: Castelán-Sánchez, H. G., & Poon, A. F. Y. (2026). Assessing the impact of phylogenetic reconstruction error on genetic reassortment inference in segmented viruses. PLOS One, 21(10), e0360852. https://doi.org/10.1371/journal.pone.0360852

Image Credits: AI Generated

DOI: 10.1371/journal.pone.0360852

Keywords: influenza A virus, reassortment, phylogenetics, H5Nx, avian influenza, subtree prune and regraft, phylogenetic error, segmented genomes, molecular evolution simulation, hemagglutinin, antigenic shift, viral evolution