AI-guided discovery of the invariant host response to viral pandemics

Debashis Sahoo1, Gajanan D Katkar2, Soni Khandelwal3

  • 1Department of Pediatrics, University of California San Diego, 9500 Gilman Drive, MC 0730, Leichtag Building 132, La Jolla, CA 92093-0831, USA; Department of Computer Science and Engineering, Jacobs School of Engineering, University of California San Diego, USA; Moores Cancer Center, University of California San Diego, USA.

Ebiomedicine
|June 15, 2021
PubMed

Insights

This study defines the host immune response in COVID-19 using AI, identifying a conserved 166-gene signature (ViP) and a 20-gene subset (severe-ViP) that predict disease severity and potential therapeutic targets.

Area of Science:

  • Immunology
  • Computational Biology
  • Virology

Background:

  • Coronavirus Disease 2019 (COVID-19) presents significant challenges to healthcare systems.
  • The "cytokine storm" is implicated in fatal COVID-19 outcomes.
  • Understanding the host immune response is crucial for managing viral pandemics.

Purpose of the Study:

  • To define the host immune response in COVID-19 using an AI-based approach.
  • To identify a gene signature associated with viral pandemics and disease severity.
  • To explore therapeutic strategies for COVID-19.

Main Methods:

  • Analysis of over 45,000 transcriptomic datasets from viral pandemics.
  • Extraction of a 166-gene signature using ACE2 as a seed gene.
  • Application of AI to analyze the gene signature's utility in disease management and therapeutics.

Main Results:

  • A 166-gene signature (ViP) was conserved across viral pandemics, with a 20-gene subset (severe-ViP) classifying disease severity.
  • Identified an IL15 cytokine storm in lung epithelial and myeloid cells; cell senescence and apoptosis correlated with severity.
  • Therapeutic goals were met in hamster models using neutralizing antibodies or EIDD-2801; IL15/IL15RA levels predicted severity in patients.

Conclusions:

  • The ViP signatures offer a framework for quantifying immune responses in viral pandemics.
  • These signatures serve as an unbiased tool for assessing disease severity and vetting drug candidates.
  • This AI-driven approach facilitates rapid assessment and therapeutic development for viral diseases.
Abstract