LAUSR

Host response to infection is a major determinant of disease severity in Ebola virus disease (EVD), but gene expression programs associated with outcome are poorly characterized. Collaborative Cross (CC) mice develop strain-dependent EVD phenotypes of differential severity, ranging from tolerance to lethality. We screen 10 CC lines and identify clinical, virologic, and transcriptomic features that distinguish tolerant from lethal outcomes. Tolerance is associated with tightly regulated induction of immune and inflammatory responses shortly following infection, as well as reduced inflammatory macrophages and increased antigen-presenting cells, B-1 cells, and γδ T cells. Lethal disease is characterized by suppressed early gene expression and reduced lymphocytes, followed by uncontrolled inflammatory signaling, leading to death. We apply machine learning to predict outcomes with 99% accuracy in mice using transcriptomic profiles. This signature predicts outcomes in a cohort of EVD patients from western Africa with 75% accuracy, demonstrating potential clinical utility.