The article introduces CellNavi, a novel deep learning framework designed to predict pivotal genes that drive complex cellular state transitions. CellNavi utilizes a gene graph-enhanced cell state manifold learned from vast, high-dimensional single-cell transcriptomics data, which integrates directional gene graph priors to capture the intrinsic features of cellular states. The framework's efficacy is demonstrated across diverse biological scenarios, including T cell differentiation, neurodegenerative disease pathogenesis, and determining the mechanism of action for drug compounds, consistently outperforming traditional network-based and differential gene expression analysis methods. The research highlights CellNavi's robust generalization capacity across various cell types and perturbation contexts, positioning it as a significant advancement for gene driver prediction and cell state manipulation.

References:

• Wang T, Pan Y, Ju F, et al. CellNavi predicts genes directing cellular transitions by learning a gene graph-enhanced cell state manifold[J]. Nature Cell Biology, 2025: 1-12.