Scholarlink

<p>Coronavirus disease 2019 (COVID-19) and other respiratory viral infections, such as influenza and respiratory syncytial virus (RSV), elicit both common and virus-specific host responses. Here, we present an integrative analysis leveraging the COVID-19 Host Genetics Initiative (HGI) GWAS data (freeze 7) and publicly available multi-omics datasets (including influenza/RSV human challenge transcriptomes and plasma proteomics) to construct an explainable AI model for comparing host infection mechanisms between COVID-19 and other viral illnesses. We identified shared antiviral pathways (type I interferon (IFN) signaling) active in host responses to all three viruses, as well as virus-specific mechanisms: for instance, SARS-CoV-2 infection induced uniquely strong coagulation and renin-angiotensin system dysregulation, along with sustained AP-1/MAPK activation, whereas influenza provoked more robust T-cell activation, and RSV triggered an excessive neutrophil-driven inflammatory response. Genetic risk pathway fingerprints from GWAS highlight that COVID-19 severity is associated with variants in IFN and inflammatory pathways, while host genetic effects in influenza point to distinct receptor usage (sialic acid biosynthesis) with minimal overlap. Mendelian randomization (MR) pinpointed key causal proteins for COVID-19 severity, including ABO (blood group glycosyltransferase) and inflammatory mediators, suggesting that host glycomic and immune factors modulate disease outcomes. Our explainable machine learning model integrated these multi-omic features to accurately distinguish COVID-19 from other viral infections, with SHAP interpretation confirming the predominance of the above mechanisms in model predictions. In summary, this cross-omics study provides a comprehensive comparative map of host responses in COVID-19 versus influenza and RSV, yielding biologically interpretable insights into both common antiviral defenses and unique pathogenic pathways. These findings inform the development of targeted therapies (IL-6 or MAPK inhibitors for COVID-19) and broad-spectrum antivirals (enhancing IFN responses) to mitigate severe respiratory viral diseases.</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

Explainable AI multiomics analysis reveals shared and divergent host responses in COVID-19 and influenza

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月27日

查看全文
  • Yan Zhang,
  • Lining Zhang,
  • Zehong Zhang,
  • Yuxi Lin,
  • Zexu Jiang,
  • Fulong Yu

摘要

Coronavirus disease 2019 (COVID-19) and other respiratory viral infections, such as influenza and respiratory syncytial virus (RSV), elicit both common and virus-specific host responses. Here, we present an integrative analysis leveraging the COVID-19 Host Genetics Initiative (HGI) GWAS data (freeze 7) and publicly available multi-omics datasets (including influenza/RSV human challenge transcriptomes and plasma proteomics) to construct an explainable AI model for comparing host infection mechanisms between COVID-19 and other viral illnesses. We identified shared antiviral pathways (type I interferon (IFN) signaling) active in host responses to all three viruses, as well as virus-specific mechanisms: for instance, SARS-CoV-2 infection induced uniquely strong coagulation and renin-angiotensin system dysregulation, along with sustained AP-1/MAPK activation, whereas influenza provoked more robust T-cell activation, and RSV triggered an excessive neutrophil-driven inflammatory response. Genetic risk pathway fingerprints from GWAS highlight that COVID-19 severity is associated with variants in IFN and inflammatory pathways, while host genetic effects in influenza point to distinct receptor usage (sialic acid biosynthesis) with minimal overlap. Mendelian randomization (MR) pinpointed key causal proteins for COVID-19 severity, including ABO (blood group glycosyltransferase) and inflammatory mediators, suggesting that host glycomic and immune factors modulate disease outcomes. Our explainable machine learning model integrated these multi-omic features to accurately distinguish COVID-19 from other viral infections, with SHAP interpretation confirming the predominance of the above mechanisms in model predictions. In summary, this cross-omics study provides a comprehensive comparative map of host responses in COVID-19 versus influenza and RSV, yielding biologically interpretable insights into both common antiviral defenses and unique pathogenic pathways. These findings inform the development of targeted therapies (IL-6 or MAPK inhibitors for COVID-19) and broad-spectrum antivirals (enhancing IFN responses) to mitigate severe respiratory viral diseases.