Vanderbilt researchers launch DUET for spatial transcriptomics
DUET predicts gene expression from histology images with 20% higher accuracy than current methods.
Deep Dive
DUET is a dual-paradigm AI framework that predicts spatially resolved gene expression from histology images. It combines parametric prediction and memory-based retrieval using single-cell references as biological constraints, achieving state-of-the-art performance on three public datasets.
Key Points
- DUET predicts gene expression from histology images with 20% higher accuracy than existing methods
- Combines parametric prediction and memory-based retrieval using single-cell references as constraints
- Achieved SOTA on three public datasets with a lightweight adapter for dynamic branch preference
Why It Matters
Enables cost-effective, high-accuracy spatial transcriptomics prediction for drug discovery and cancer research.