A New Way to Trace SARS-CoV-2 Variants Through Weighted Network Analysis of Frequency Trajectories of Mutations

Qiang Huang, Qiang Zhang, Paul W. Bible, Qiaoxing Liang, Fangfang Zheng, Ying Wang, Yuantao Hao, Yu LiuFrontiers in Microbiology 2022 · DOI

TL;DR

This paper introduces a novel weighted network analysis framework to track SARS-CoV-2 variants in near real-time by analyzing the frequency trajectories of mutations (FTMs). This method bypasses the delays associated with traditional subtype assignment, enabling earlier detection and more effective public health responses to emerging variants.

What Problem Does This Paper Address?

The COVID-19 pandemic highlighted the critical need for rapid identification and tracking of SARS-CoV-2 variants. Existing surveillance methods, such as those used by GISAID, Pangolin, and CMM, often rely on pre-defined genetic features and lag times for subtype assignment. This can delay the detection of new or evolving variants, hindering timely public health interventions. Furthermore, some methods offer either too detailed or too broad categorization, impacting the efficiency of risk monitoring and communication. The authors aim to address this by developing a method that can identify variants dynamically and in real-time, without requiring prior knowledge of their lineage.

Methods

The core of this paper is a weighted network framework for variant tracing. Here's a breakdown of the key steps:

  1. Sequence Curation: SARS-CoV-2 genomes are retrieved from GISAID, filtered for quality and completeness.
  2. Bioinformatic Analysis for Mutation Calling: Single nucleotide polymorphisms (SNPs) and insertions/deletions (INDELs) are identified using a bioinformatic pipeline.
  3. Generation and Filtration of FTMs: Frequency trajectories of mutations (FTMs) are generated for each mutation site over time. Mutations with consistently low frequencies are filtered out to focus on significant trends.
  4. Weighted Network Construction:
    • Each mutation's FTM is treated as a node in a network.
  5. The Pearson correlation coefficient is used to quantify the synchronous relationship between pairs of FTMs.
  6. These correlations are transformed into connection strengths using a power function, creating a weighted network where stronger connections indicate more synchronized mutation patterns.
  7. Module Identification: Hierarchical clustering based on topological overlap (a measure of interconnectedness considering shared neighbors) is used to group tightly connected mutations into "modules." These modules are hypothesized to represent distinct variants.
  8. Core Mutations for Variant Determination: Within each module, "core" mutations with high connection strengths are identified. These core mutations are used to define and characterize a specific variant.
  9. Phylogenetic Branch Assessment: The identified variants are validated against global phylogenetic trees to confirm their lineage and evolutionary relationships.

This approach is novel in its direct use of FTMs as network nodes and its reliance on weighted network analysis and topological overlap for module identification, bypassing the need for pre-assigned lineages.

Datasets

  • Worldwide GISAID data: Used for benchmarking the methodology and demonstrating its capability in identifying global pandemic variants.
  • Regional data from India and South Africa: Used to showcase the framework's application in real-world variant surveillance and its ability to elucidate national dispersal histories and co-circulation patterns.

The paper does not introduce a new benchmark dataset but rather applies its method to existing, large-scale genomic data.

Key Insights and Ideas

  • Synchronous FTMs as Variant Signatures: The fundamental insight is that mutations that rise and fall in frequency together are likely to be part of the same evolving variant. This temporal co-occurrence is a powerful indicator.
  • Network Modules as Variants: Grouping these synchronous FTMs into network modules provides a data-driven way to define variants without relying on pre-existing classification systems.
  • Early Warning System Potential: The method's ability to detect emerging patterns in FTMs before they become dominant suggests its potential as an early warning system for new variants.
  • Bridging the Gap in Surveillance: The framework offers a middle ground between overly broad (e.g., GISAID) and overly granular (e.g., Pangolin) surveillance methods, providing a more actionable level of detail.
  • Inclusion of Indels and Recombinations: The method's ability to incorporate insertion/deletion events, not just substitutions, is a significant advantage over methods that primarily focus on SNPs.

Key Results

  • Global Variant Identification: The weighted network framework successfully identified major SARS-CoV-2 variants from worldwide GISAID data, showing high specificity and sensitivity when mapped against global phylogenetic trees.
  • Earlier Detection in Regional Data:
    • In India, the method provided an earlier warning for Delta (B.1.617.2) than its official reporting as a Variant Under Investigation (VUI) by the WHO.
  • In South Africa, the framework tracked dominant variants like Omicron (B.1.1.529) and Beta (B.1.351) earlier than their official VUI/VOC dates.
  • Elucidation of Dispersal History: The analysis of regional data clearly mapped the national dispersal history and co-circulation patterns of viral variants.
  • Identification of Sub-variants: The method was able to identify more detailed subtypes and descendants of major variants compared to existing classification systems like CMM and Pangolin.
  • Modular Representation of Epidemic Stages: Module clusters within the network effectively represented different epidemic stages and the contemporaneous variants circulating during those times.

Limitations and Future Work

  • Nominated Lineage vs. Perceptive Insights: The mutation modules detected by the workflow may not always directly correspond to a formally nominated lineage but offer valuable insights into novel variants with potentially higher transmission.
  • Assumption of Independence: The assumption of independence between FTMs at the same nucleotide sites might not always hold true, though the authors suggest it doesn't significantly impact results.
  • Empirical Thresholds: The filtration threshold for FTMs is chosen empirically, which could lead to the loss of less frequent but potentially important variants. The authors suggest this is a trade-off between detectability and discriminability.
  • Future Work: The authors suggest that as more samples become available, analysis at a higher spatial resolution could be beneficial. They also emphasize the need for timely availability of viral genomes for effective surveillance.

Summary updated

Abstract

Early detection of SARS-CoV-2 variants enables timely tracking of clinically important strains in order to inform the public health response. Current subtype-based variant surveillance depending on prior subtype assignment according to lag features and their continuous risk assessment may delay this process. We proposed a weighted network framework to model the frequency trajectories of mutations (FTMs) for SARS-CoV-2 variant tracing, without requiring prior subtype assignment. This framework modularizes the FTMs and conglomerates synchronous FTMs together to represent the variants. It also generates module clusters to unveil the epidemic stages and their contemporaneous variants. Eventually, the module-based variants are assessed by phylogenetic tree through sub-sampling to facilitate communication and control of the epidemic. This process was benchmarked using worldwide GISAID data, which not only demonstrated all the methodology features but also showed the module-based variant identification had highly specific and sensitive mapping with the global phylogenetic tree. When applying this process to regional data like India and South Africa for SARS-CoV-2 variant surveillance, the approach clearly elucidated the national dispersal history of the viral variants and their co-circulation pattern, and provided much earlier warning of Beta (B.1.351), Delta (B.1.617.2), and Omicron (B.1.1.529). In summary, our work showed that the weighted network modeling of FTMs enables us to rapidly and easily track down SARS-CoV-2 variants overcoming prior viral subtyping with lag features, accelerating the understanding and surveillance of COVID-19.
BibTeX
@inproceedings{huang2022new,
  author = {Qiang Huang and Qiang Zhang and Paul W. Bible and Qiaoxing Liang and Fangfang Zheng and Ying Wang and Yuantao Hao and Yu Liu},
  title = {{A New Way to Trace SARS-CoV-2 Variants Through Weighted Network Analysis of Frequency Trajectories of Mutations}},
  booktitle = {Frontiers in Microbiology},
  year = {2022},
  doi = {10.3389/fmicb.2022.859241},
}
Figure