Multiomics and Explainable AI Reveal Shared and Distinct Host Responses in COVID-19 and Influenza

Overview

This study integrates genetic, transcriptomic, proteomic, and immune repertoire data to delineate common and virus-specific host responses in COVID-19 and influenza. Using explainable AI, it identifies conserved interferon-driven antiviral defenses alongside COVID-19-specific thrombo-inflammatory and endothelial dysfunction signatures.

Background

COVID-19 severity varies widely due to host genetic and immune regulatory differences, with key loci implicated in interferon signaling and inflammatory pathways. While COVID-19 shares pathophysiological features with influenza and RSV, it uniquely exhibits pulmonary microthrombosis and endotheliopathy. Prior studies have shown overlapping and divergent immune signatures across these viruses, but comprehensive multi-omics comparisons anchored in host genetics have been lacking. This study leverages public multi-omic datasets and genetic association data to dissect shared versus distinct antiviral mechanisms.

Data Highlights

Key Findings

• Host genetic loci for COVID-19 implicate innate antiviral defense (OAS1/2/3, IFNAR2) and inflammatory pathways (TYK2, ABO).
• COVID-19 exhibits a dysregulated immune profile with blunted systemic type I IFN but elevated proinflammatory cytokines (IL-6, TNF) and maladaptive myeloid responses.
• Multi-omic integration reveals persistent AP-1/MAPK and thrombo-inflammatory network activation unique to severe COVID-19 compared to influenza.
• Explainable AI classifiers distinguish COVID-19 from influenza based on pathway-level signatures, highlighting shared IFN backbones and COVID-19-specific endothelial/coagulation involvement.
• Proteome-wide Mendelian randomization nominates cytokine and endothelial-coagulation mediators as causal drivers of severe COVID-19, with limited overlap in influenza risk loci.
• Comparative analyses with RSV underscore universal antiviral modules versus virus-specific immune adaptations.

Clinical Implications

These findings support timing-specific host-directed interventions such as interferon augmentation early in infection and modulation of IL-6 and endothelial/coagulation pathways in severe COVID-19. Understanding distinct immune signatures can guide personalized therapeutic strategies and improve management of viral respiratory diseases. The genetic anchoring of multi-omic data enhances confidence in targeting key pathways.

Conclusion

Integrating genetics with multi-omics and explainable AI reveals a shared antiviral interferon framework alongside COVID-19-specific thrombo-inflammatory and endothelial dysfunction signatures. This comprehensive approach advances mechanistic understanding and prioritizes host-directed therapeutic hypotheses.

References

1. COVID-19 Host Genetics Initiative (HGI) -- Genome-wide association studies
2. COMBAT Consortium -- Multi-Omic Blood Atlas of COVID-19
3. GEO GSE73072 -- Controlled Human Viral Challenge Transcriptomics
4. Mendelian Randomization Studies -- Cytokine and Endothelial Mediators in COVID-19