Predicting Influenza A Virus Infection in the Lung from Hematological Data with Machine Learning.

Area of Science:

• * Virology
• * Immunology
• * Computational Biology

Background:

• * Monitoring respiratory infections like influenza A virus (IAV) often requires invasive and costly methods.
• * Limited non-invasive tools exist to quantitatively assess host immune responses and pathogen burden in the lungs.
• * Objective evaluation of disease progression in respiratory infections remains a challenge.

Purpose of the Study:

• * To develop and validate machine learning models for predicting lung viral burden and immune markers from hematological data.
• * To identify key hematological parameters indicative of respiratory infection status.
• * To establish a minimally invasive approach for monitoring influenza virus infections.

Main Methods:

• * Development and testing of supervised machine learning models.
• * Utilization of a standardized murine model of respiratory influenza A virus (IAV) infection.
• * Independent in vivo experiments for data acquisition, training, and model validation.

Main Results:

• * Hematological data accurately predicted lung viral load, neutrophil counts, and cytokine levels (e.g., IFN-γ, IL-6).
• * Machine learning models identified blood granulocytes and platelets as crucial predictors of IAV infection.
• * The study demonstrated the feasibility of inferring lung infection status from peripheral blood parameters.

Conclusions:

• * Minimally invasive hematological parameters can serve as reliable surrogates for monitoring lung viral burden and immune responses in IAV infections.
• * The developed in silico tools offer a non-invasive strategy for improved tracking and management of influenza and potentially other respiratory infections.
• * This approach provides a novel, non-invasive perspective on lung disease processes.