Uncovering Zika virus-specific neurodevelopmental disruptions through deep learning and transcriptomic profiling of human neural cells

Zika virus (ZIKV) infection during pregnancy can cause severe brain defects in newborns, yet the exact changes it triggers in human neural cells remain unclear. This study combined gene expression profiling with interpretable deep learning to uncover how ZIKV alters key genetic programs in neural progenitor cells (NPCs). We trained a neural network that accurately distinguished infected from healthy cells and used tools like SHAP and integrated gradients to identify important genes driving this distinction. Among these, NEAT1, MALAT1, GAPDH, and XIST emerged as central regulators. Further analyses showed that ZIKV disrupts genes tied to cell division, stress responses, and brain development. Unlike infections by related viruses like Dengue and West Nile, ZIKV activated a unique set of genes. Testing our model on an independent dataset confirmed its reliability. These highlight specific genetic signatures of ZIKV infection, which offers possible targets for early diagnosis or new treatments against virus-induced brain disorders.