Analyzing the relationship between gene expression and phenotype in space-flown mice using a causal inference machine learning ensemble

Spaceflight has several detrimental effects on human and rodent health. For example, liver dysfunction is a common phenotype observed in space-flown rodents, and this dysfunction is partially reflected in transcriptomic changes. Studies linking transcriptomics with liver dysfunction rely on tools which exploit correlation, but these tools make no attempt to disambiguate true correlations from spurious ones. In this work, we use a machine learning ensemble of causal inference methods called the Causal Research and Inference Search Platform (CRISP) which was developed to predict causal features of a binary response variable from high-dimensional input. We used CRISP to identify genes robustly correlated with a lipid density phenotype using transcriptomic and histological data from the NASA Open Science Data Repository (OSDR). Our approach identified genes and molecular targets not predicted by previous traditional differential gene expression analyses. These genes are likely to play a pivotal role in the liver dysfunction observed in space-flown rodents, and this work opens the door to identifying novel countermeasures for space travel.