32nd International Dynamics & Evolution of Human Viruses

Phylogenetic Discordance Cannot Identify Genomic Reassortment Events: A Case Study with the Avian Influenza Virus H5Nx

Not scheduled

20m

Abbaye de Royaumont, Asnières-sur-Oise, France

Oral Genomics & bioinformatics

Speaker

Dr Hugo Castelan (Department of Pathology and Laboratory Medicine, Western University, London, ON, Canada)

Description

Reassortment is the exchange of genomic segments between viruses. It is a significant feature of the evolution of influenza A viruses (IAV), where it can yield new combinations between highly divergent lineages (antigenic shift). In this study, we evaluated the reliability of standard comparative methods for reconstructing reassortment events in the evolutionary history of H5Nx IAV genomes. We retrieved a total of 11,765 complete H5Nx genome sequences from NCBI Genbank and GISAID. These data were screened for duplicate records and then segment-specific alignments were generated with MAFFT. We used IQ-TREE to reconstruct a maximum likelihood (ML) tree for the alignment of HA sequences, and then progressively removed any terminal with a length exceeding 0.05. Using this reduced set of divergent genomes, we reconstructed ML trees from
each segment. We generated optimized tanglegrams and maximum agreement forests, i.e., counting subtree-prune-regraft (SPR) events, to quantify phylogenetic incongruence between segments.
Next, we used Pyvolve to simulate sequence evolution along the HA tree for each segment, scaling tree lengths to the observed levels of diversity for each segment. Because these simulations were run on the same tree, there was a complete absence of reassortment. Nevertheless, we observed no significant difference in entanglement values for tanglegrams (range 0.07−0.3, paired Wilcoxon P = 0.34) nor the numbers of SPR events (range143−168, P = 0.4) between trees for real or simulated datasets. This implies a high false positive rate due to uncertainty in reconstructing trees.
In addition, we found no significant difference between real and simulated data in the proportion of SPRs involving single branches (overall 87.3%, P= 0.078), or the phylogenetic distance between "regraft points" for SPRs between trees. These results imply that there are no simple quantitative criteria that identify actual reassortment events from phylogenetic discordance, and suggest that studies employing similar methods may have overestimated the amount of reassortment in IAV genomes.