.png){kind=icon}
.png){kind=icon}
Return
GenAR: Next-scale autoregressive generation for spatial gene expression prediction
.png){kind=icon}
PRE
AI
DOI：10.1016/j.media.2026.104232.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• GenAR predicts spatial gene expression from H&E + coordinates via next-scale autoregression. • Hierarchical gene groups enable coarse-to-fine decoding and cross-gene dependence. • Discrete count-token generation predicts raw integer counts on the physical scale. • State-of-the-art on five spatial transcriptomics datasets across diverse tissues. • Improves zero recovery and count accuracy with ∼6.4× fewer FLOPs than diffusion.
Keywords:
Histopathology
Spatial transcriptomics
Gene expression prediction
Autoregressive
Journal
IF:
11.8
Papers:
3.8K
Citations:
2.4W