Digital Biomarker Models for Prediction of Infectious Disease Susceptibility

Abstract

Acute respiratory viral infection (ARVI) represents one of the most prevalent infectious diseases affecting mankind. With the threat of COVID-19 still looming over us, we have witnessed the substantial threat ARVI poses to world health and economy, extinguishing millions of lives and costing trillions of dollars. This sets the context for the research of this thesis: using digital biomarkers to distinguish between individuals who are susceptible to becoming severely infected and/or infectious before an infection is clinically detectable. The development of such biomarkers can have both clinical and epidemiological impact in terms of identifying individuals who are either vulnerable to severe infection or those who may become highly infectious. The digital biomarkers and associated analysis methods are developed and validated on longitudinal data collected by our clinical collaborators from two different ARVI challenge studies. The first study provides data on healthy human volunteers who were inoculated with the common cold and the second study provides data on volunteers inoculated with the flu. Digital biomarkers include molecular, physiological and cognitive data continuously collected from blood, wearable devices and cognitive testing of the study participants. The findings of our research on digitally measurable susceptibility factors are wide-ranging. We find that circadian rhythm at the molecular scale (biochronicity) plays an important role in mediating both the susceptibility and the response to severe infection, revealing groups of gene expression markers that differentiate the responses of low infected and high infected individuals. Using a high dimensional representation of physiological signals from a wearable device, we find that an infection response and its onset time can be reliably predicted at least 24 hours before peak infection time. We find that a certain measure of variability in pre-exposure cognitive function is highly associated with the post-exposure severity of infection.