Deep Cross-Organism Generalization of the Physiological Effects of Spaceflight from Mammalian Model Organisms to Humans

Abstract The forthcoming human deep space exploration missions necessitate a thorough understanding of the impact of spaceflight conditions on human physiological systems. The NASA Open Science Data Repository (OSDR; https://osdr.nasa.gov/bio/ ) serves as a valuable resource, housing data derived from model organisms and human experiments conducted in spaceflight and terrestrial microgravity analogues. Machine Learning applications could maximize the use of existing data to understand and ultimately counteract physiological abnormalities during long-term missions. In our present study, we identified enriched terms and pathways associated with significantly dysregulated genes within each species and across orthologous counterparts. We also generated AI-ready merged meta-datasets comprised of musculoskeletal tissues from Mus musculus and Homo sapiens organisms. We then applied a series of supervised Machine Learning models to classify genes that were significantly over-expressed and under-expressed. Subsequently, we explored the utility of Transfer Learning in this domain by pretraining a model on the larger Mus musculus merged dataset and then refining it on the smaller Homo sapiens dataset. This approach showcases the potential of Transfer Learning in providing an insight into the effective transfer of information from model organisms to humans, offering a robust framework for advancing research in space biology and developing countermeasures for long-duration space exploration.