Abstract
To explore the differences and relationships between the available SARS-CoV-2 strains and predict the potential evolutionary direction of these strains, we employ hierarchical clustering analysis to investigate the evolutionary relationships between the SARS-CoV-2 strains utilizing genomic sequences collected in China until January 7, 2023. We encode the sequences of the existing SARS-CoV-2 strains into numerical data through the k-mer algorithm, then propose four methods to select the representative sample from each type of strain to comprise the dataset for clustering analysis. Three hierarchical clustering algorithms named Ward-Euclidean, Ward-Jaccard, and Average-Euclidean are introduced through combining Euclidean and Jaccard distance with the Ward and Average linkage clustering algorithms. Experimental results reveal that BF.28, BE.1.1.1, BA.5.3, and BA.5.6.4 strains exhibit distinct characteristics which are not observed in other types of SARS-CoV-2 strains, suggesting they are the majority potential sources for future SARS-CoV-2 evolution. Moreover, BA.2.75, CH.1.1, BA.2, BA.5.1.3, BF.7, and B.1.1.214 strains demonstrate enhanced abilities in terms of immune evasion, transmissibility, and pathogenicity.
Introduction
As SARS-CoV-2 continues to circulate globally, its rapid mutation rate creates a "moving target" for public health interventions. While traditional phylogenetic trees are the gold standard for tracking ancestry, the sheer volume of genomic data requires efficient, scalable clustering methods to identify broader evolutionary trends. This study addresses the challenge of systematically categorizing diverse strains to identify relationships and predict which "potential strains" might emerge next, focusing specifically on the critical period in China leading up to early 2023.
Methodology
The researchers developed a multi-step pipeline to transform raw genetic sequences into actionable evolutionary maps.
Feature Encoding via k-mer
To make DNA sequences machine-readable, the $k$-mer algorithm was employed. This process breaks down the genetic code into overlapping subsequences of length $k$ (where $k=1$ to $5$). This transformation turned each viral sequence into a numerical vector of 1,364 features, representing the frequency of specific nucleotide combinations.
Representative Sampling and Clustering
With 172 different strain types identified, the authors proposed four methods to select a "representative sample" (centroid) for each strain type to ensure computational efficiency. They then applied three hierarchical clustering combinations: Ward-Euclidean, Ward-Jaccard, and Average-Euclidean. These methods allow for different perspectives on viral similarity, with Euclidean distance focusing on feature magnitude and Jaccard highlighting the presence or absence of specific mutations.
Results and Discussion
The analysis successfully grouped the 172 lineages into six distinct clusters. The Ward-Jaccard method was particularly effective at separating the Delta variant (e.g., AY.57) from Omicron variants, underscoring the significant genetic leap between these two Variants of Concern.
Evolutionary Hubs and Functional Clustering
The study identified BF.28, BE.1.1.1, BA.5.3, and BA.5.6.4 as strains with distinct genetic signatures, positioning them as the primary potential sources for future evolution. Furthermore, strains such as BA.2.75, CH.1.1, and BF.7 clustered together based on shared enhancements in immune evasion and pathogenicity. This suggests that the mathematical clustering reflects real-world biological advantages driven by specific spike protein mutations like L452R and F486V.
Conclusion
By combining $k$-mer sequence encoding with hierarchical clustering, this framework provides a high-level map of the SARS-CoV-2 evolutionary landscape. Closely monitoring the identified high-frequency clusters and "hub" strains is essential for mitigating the impact of future variants on public health. Future work will involve expanding the geographic scope of the data and conducting molecular-level investigations to confirm the biological mechanisms behind these evolutionary paths.
Summary updated