Scholarlink

<p>SARS-CoV-2 exhibits substantial genomic diversity, with emerging variants posing challenges to global public health. In this study, we applied unsupervised clustering to genome sequences using k-mer encoding to explore viral genetic variation. Both 2-mer and 3-mer representations were analyzed with hierarchical clustering and dimensionality reduction techniques (PCA and t-SNE) to identify underlying genomic structure. Internal validation metrics consistently indicated four optimal clusters, with the 3-mer + t-SNE configuration achieving the highest cluster quality (Silhouette Score = 0.648, Calinski-Harabasz Index = 2443.3, Davies-Bouldin Index = 0.325). Cluster analysis revealed distinct lineage-defining mutations, including D614G and P681H in Clusters 1 and 4, and N501Y and E484A in Cluster 4, highlighting their functional relevance in infectivity and immune escape. Sensitivity analysis confirmed that these mutations significantly contributed to k-mer feature vectors and cluster separation. External validation using PANGO lineage annotations demonstrated strong concordance (ARI = 0.83, NMI = 0.86), confirming that the clustering framework effectively recapitulates known viral population structure. These results provide a scalable and biologically meaningful approach for monitoring SARS-CoV-2 genomic diversity, detecting emerging variants, and supporting genomic epidemiology efforts.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Integrative PCA–t-SNE clustering framework for comparative genomic analysis of regional SARS-CoV-2 variants

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月3日

查看全文
  • S. Umamaheswari,
  • K. Hemalatha,
  • P. Thamizhazhagan,
  • A. B. Feroz Khan

摘要

SARS-CoV-2 exhibits substantial genomic diversity, with emerging variants posing challenges to global public health. In this study, we applied unsupervised clustering to genome sequences using k-mer encoding to explore viral genetic variation. Both 2-mer and 3-mer representations were analyzed with hierarchical clustering and dimensionality reduction techniques (PCA and t-SNE) to identify underlying genomic structure. Internal validation metrics consistently indicated four optimal clusters, with the 3-mer + t-SNE configuration achieving the highest cluster quality (Silhouette Score = 0.648, Calinski-Harabasz Index = 2443.3, Davies-Bouldin Index = 0.325). Cluster analysis revealed distinct lineage-defining mutations, including D614G and P681H in Clusters 1 and 4, and N501Y and E484A in Cluster 4, highlighting their functional relevance in infectivity and immune escape. Sensitivity analysis confirmed that these mutations significantly contributed to k-mer feature vectors and cluster separation. External validation using PANGO lineage annotations demonstrated strong concordance (ARI = 0.83, NMI = 0.86), confirming that the clustering framework effectively recapitulates known viral population structure. These results provide a scalable and biologically meaningful approach for monitoring SARS-CoV-2 genomic diversity, detecting emerging variants, and supporting genomic epidemiology efforts.