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

bioRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Feb. 8, 2024

Abstract Public health researchers and practitioners commonly infer phylogenies from viral genome sequences to understand transmission dynamics identify clusters of genetically-related samples. However, viruses that reassort or recombine violate phylogenetic assumptions require more sophisticated methods. Even when are appropriate, they can be unnecessary difficult interpret without specialty knowledge. For example, pairwise distances between enough related samples assign new existing clusters. In this work, we tested whether dimensionality reduction methods could capture known genetic groups within two human pathogenic cause substantial morbidity mortality frequently recombine, respectively: seasonal influenza A/H3N2 SARS-CoV-2. We applied principal component analysis (PCA), multidimensional scaling (MDS), t-distributed stochastic neighbor embedding (t-SNE), uniform manifold approximation projection (UMAP) with well-defined clades either reassortment (H3N2) recombination (SARS-CoV-2). each low-dimensional sequences, calculated the correlation Euclidean in a hierarchical clustering method embedding. measured accuracy compared previously defined clades, clusters, recombinant lineages. found MDS embeddings accurately represented including intermediate placement SARS-CoV-2 lineages parental Clusters t-SNE recapitulated H3N2 groups, show simple statistical biological model represent relationships for relevant viruses. Our open source implementation these easily inappropriate.

Language: Английский

---

Deep learning of SARS-CoV-2 outbreak phylodynamics with contact tracing data

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: June 11, 2024

Abstract Deep learning has emerged as a powerful tool for phylodynamic analysis, addressing common computational limitations affecting existing methods. However, notable disparities exist between simulated phylogenetic trees used training deep models and those derived from real-world sequence data, necessitating thorough examination of their practicality. We conducted comprehensive evaluation model performance by assessing an inference phylodynamics, PhyloDeep, against realistic characterized SARS-CoV-2. Our study reveals the poor predictive accuracy PhyloDeep trained on when applied to data. Conversely, demonstrate improved predictions, despite not being infallible, especially in scenarios where superspreading dynamics are challenging capture accurately. Consequently, we find markedly through integration minimal contact tracing Applying this approach sample SARS-CoV-2 sequences partially matched Hong Kong yields informative estimates potential beyond scope data alone. findings enhancing processing low resolution complementary integration, ultimately increasing precision epidemiological predictions crucial public health decision making outbreak control.

Language: Английский

---

Infections are not alike: the effects of covariation between individual susceptibility and transmissibility on epidemic dynamics

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 12, 2024

Abstract Individual-level variation in susceptibility to infection and transmissibility of can affect population-level dynamics epidemic outbreaks. Prior work has incorporated independent or individuals into compartmental models. Here, we develop assess a mathematical framework that includes covariation transmissibility. We show uncorrelated leads an effective distribution constant coefficient such the match those with alone, providing baseline for comparison across different correlation structures. Increasing between increases both speed strength outbreak – is indicative outbreaks which might be strongly structured by contact rate variation. In contrast, negative correlations lead overall weaker caveat transmission over time. either case, shift distribution, thereby modifying as susceptible population depleted. Overall, this demonstrates how (often unaccounted) shape course final sizes. Highlights Developed models incorporating covariation. Identified eigendistributions force infection. Uncorrelated reduces alone. Positive basic reproduction number. give faster, stronger, more likely Effective rates increase time correlations.

Language: Английский

---