Dimensionality reduction distills complex evolutionary relationships in seasonal influenza and SARS-CoV-2

Abstract

Public health researchers and practitioners commonly infer phylogenies from viral genome sequences to understand transmission dynamics and identify clusters of genetically-related samples. However, viruses that reassort or recombine violate phylogenetic assumptions and require more sophisticated methods. Even when phylogenies are appropriate, they can be unnecessary or difficult to interpret without specialty knowledge. For example, pairwise distances between sequences can be enough to identify clusters of related samples or assign new samples to existing phylogenetic clusters. In this work, we tested whether dimensionality reduction methods could capture known genetic groups within two human pathogenic viruses that cause substantial human morbidity and mortality and frequently reassort or recombine, respectively: seasonal influenza A/H3N2 and SARS-CoV-2. We applied principal component analysis (PCA), multidimensional scaling (MDS), t-distributed stochastic neighbor embedding (t-SNE), and uniform manifold approximation and projection (UMAP) to sequences with well-defined phylogenetic clades and either reassortment (H3N2) or recombination (SARS-CoV-2). For each low-dimensional embedding of sequences, we calculated the correlation between pairwise genetic and Euclidean distances in the embedding and applied a hierarchical clustering method to identify clusters in the embedding. We measured the accuracy of clusters compared to previously defined phylogenetic clades, reassortment clusters, or recombinant lineages. We found that MDS maintained the strongest correlation between pairwise genetic and Euclidean distances between sequences and best captured the intermediate placement of recombinant lineages between parental lineages. Clusters from t-SNE most accurately recapitulated known phylogenetic clades and recombinant lineages. Both MDS and t-SNE accurately identified reassortment groups. We show that simple statistical methods without a biological model can accurately represent known genetic relationships for relevant human pathogenic viruses. Our open source implementation of these methods for analysis of viral genome sequences can be easily applied when phylogenetic methods are either unnecessary or inappropriate.Author summaryTo track the progress of viral epidemics, public health researchers often need to identify groups of genetically-related samples. A common approach to find these groups involves inferring the complete evolutionary history of virus samples using phylogenetic methods. However, these methods assume that new viruses descend from a single parent, while many viruses including seasonal influenza and SARS-CoV-2 produce offspring through a form of sexual reproduction that violates this assumption. Additionally, phylogenies may be unnecessarily complex or unintuitive when researchers only need to find and visualize clusters of related samples. We tested an alternative approach by applying widely-used statistical methods (PCA, MDS, t-SNE, and UMAP) to create 2- or 3-dimensional maps of virus samples from their pairwise genetic distances and identify clusters of samples that place close together in these maps. We found that these statistical methods without an underlying biological model could accurately capture known genetic relationships in populations of seasonal influenza and SARS-CoV-2 even in the presence of sexual reproduction. The conceptual and practical simplicity of our open source implementation of these methods enables researchers to visualize and compare human pathogenic virus samples when phylogenetic methods are unnecessary or inappropriate.