Machine Learning Ensemble Reveals Distinct Molecular Pathways of Retinal Damage in Spaceflown Mice
Journal:
bioRxiv
Published Date:
Jan 29, 2026
Abstract
Spaceflight-associated neuro-ocular syndrome (SANS) threatens astronaut health during long-duration missions, yet its molecular pathology remains unclear. Two physiological pathways exist: one attributes SANS to microgravity-induced cephalic fluid shift causing increased intracranial pressure, altered CSF flow, blood-brain barrier disruption, and optic disc edema; the other emphasizes radiation-induced oxidative stress affecting electrolyte channels. However, the exact mechanisms are not completely understood in this unique pathophysiologic barrier to human spaceflight. We investigated two cellular stress markers in the retinal tissues of mice (n=16) flown on the International Space Station (ISS): 4-hydroxynonenal (4-HNE), indicating oxidative stress and lipid peroxidation, and DNA fragmentation detected via TUNEL assay, indicating apoptosis. Using an ensemble of five machine learning algorithms to analyze gene expression profiles, we identified genes that are most predictive of 4-HNE and TUNEL phenotype values. By identifying the most predictive protein-coding genes, we identified associated molecular pathways whose dysregulation may mechanistically link spaceflight conditions to SANS-related retinal damage. Our findings support oxidative stress and DNA damage as primary drivers in SANS pathology. This computational approach uses robust correlation to reveal pathway associations that may inform future therapeutic investigations, advancing our understanding of SANS molecular mechanisms, and offering potential intervention strategies for protecting astronaut vision during future deep-space missions.
