Spatial transcriptomics enables the study of tissue organization in situ, but many high-resolution platforms measure only a limited gene panel, leaving much of the transcriptome unobserved. Although deep learning methods can reconstruct missing genes from matched single-cell references, downstream differential expression (DE) analysis remains unreliable because prediction uncertainty and spatially structured sources of variation are typically ignored. These factors can bias effect estimates and inflate false discoveries. We present TIDEST, a framework for DE testing after spatial transcriptomic imputation. TIDEST uses information from measured genes to correct systematic errors in reconstructed expression and adjusts for latent spatial variation, such as tissue architecture or cell type composition, that can create spurious differences between biological groups. Across extensive simulations, TIDEST maintains substantially better error control than existing approaches while preserving power. Applications to mouse brain, human glioblastoma, and human breast cancer data recover biologically meaningful DE signals that are missed or distorted by conventional analyses. TIDEST provides a principled framework for DE analysis on reconstructed spatial transcriptomes.
Roeder, K., Lei, J., Testa, L.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 5
- Comments 0