Multimodal foundation model predicting transcriptome-wide virtual spatial transcriptomics from histology.
-
Updated
Nov 16, 2026 - Python
Multimodal foundation model predicting transcriptome-wide virtual spatial transcriptomics from histology.
Code for "Probabilistic uncertainty in virtual spatial transcriptomics from histology is largely redundant with the predicted mean" (bioRxiv 2026). Benchmark of aleatoric and epistemic uncertainty (Gaussian/Poisson/NB heads, deep ensembles, MC dropout, Laplace) for gene expression predicted from H&E on 10x Visium.
Analysis code for "Virtual spatial transcriptomics links kidney histopathology to disease-associated molecular programs": DeepSpot-M predicts 19,338 genes from H&E whole-slide images of 1,534 KPMP kidney biopsies, validated with Visium, Xenium and rat kidney injury data.
To associate your repository with the virtual-spatial-transcriptomics topic, visit your repo's landing page and select "manage topics."