Probabilistic uncertainty in virtual spatial transcriptomics from histology is largely redundant with the predicted mean
- 1ETH Zurich
Abstract
Deep learning has enabled virtual spatial transcriptomics from histology, offering a scalable approach to molecular profiling of routinely archived tissue without molecular assays. However, reliable uncertainty estimates are essential for identifying potentially inaccurate predictions, and it remains unclear whether existing methods provide useful information about prediction reliability. Here, we present a systematic benchmark of uncertainty estimation for virtual spatial transcriptomics across three cancer cohorts spanning melanoma, lung, and kidney tumors. We evaluate aleatoric and epistemic uncertainty using probabilistic likelihoods and approximate Bayesian methods alongside histology-free and training-free baselines, assessing both calibration and error informativeness. Across cohorts and methods, uncertainty is associated with prediction error, but the predicted mean largely explains this association. After accounting for the mean, uncertainty provides only weak information about larger errors. Moreover, good calibration does not imply informative uncertainty, with simple distributions fitted without histology achieving comparable calibration. We therefore distinguish calibrated from informative uncertainty and argue that uncertainty should provide information about prediction error beyond the predicted mean. This benchmark provides a framework for systematic uncertainty evaluation in virtual spatial transcriptomics. Code availability: https://github.com/ratschlab/uncertainty-vst-2026
Related
- HESTIA: Scalable Multimodal Integration of Histology and High-Resolution Spatial Transcriptomics for Robust Spatial Domain Identification
- VGL-stGen: Learning a shared multimodal model for lung histology and spatial transcriptomics
- SpaFactor: Lightweight Spatial Context-Aware Gene Program Modeling for Histology-to-Transcriptomics Inference
- Quantitative Donor Histology and Transcriptomics Link Pancreatic Composition and Precancers Across the Adult Lifespan
- Uncertainty-aware AI for tumor subtyping with histology and immunohistochemistry: A multi-center study in Renal Cell Carcinoma
BibTeX
@misc{batique2026probabilistic,
title = {{Probabilistic uncertainty in virtual spatial transcriptomics from histology is largely redundant with the predicted mean}},
author = {T. Batique and K. Nonchev and V. H. Koelzer and G. Ratsch},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.23.760875},
url = {https://www.biorxiv.org/content/10.64898/2026.10.23.760875v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Probabilistic uncertainty in virtual spatial transcriptomics from histology is largely redundant with the predicted mean
AU - T. Batique
AU - K. Nonchev
AU - V. H. Koelzer
AU - G. Ratsch
PY - 2026
DA - 2026/10/23
PB - bioRxiv
DO - 10.64898/2026.10.23.760875
UR - https://www.biorxiv.org/content/10.64898/2026.10.23.760875v1
AB - Deep learning has enabled virtual spatial transcriptomics from histology, offering a scalable approach to molecular profiling of routinely archived tissue without molecular assays. However, reliable uncertainty estimates are essential for identifying potentially inaccurate predictions, and it remains unclear whether existing methods provide useful information about prediction reliability. Here, we present a systematic benchmark of uncertainty estimation for virtual spatial transcriptomics across three cancer cohorts spanning melanoma, lung, and kidney tumors. We evaluate aleatoric and epistemic uncertainty using probabilistic likelihoods and approximate Bayesian methods alongside histology-free and training-free baselines, assessing both calibration and error informativeness. Across cohorts and methods, uncertainty is associated with prediction error, but the predicted mean largely explains this association. After accounting for the mean, uncertainty provides only weak information about larger errors. Moreover, good calibration does not imply informative uncertainty, with simple distributions fitted without histology achieving comparable calibration. We therefore distinguish calibrated from informative uncertainty and argue that uncertainty should provide information about prediction error beyond the predicted mean. This benchmark provides a framework for systematic uncertainty evaluation in virtual spatial transcriptomics. Code availability: https://github.com/ratschlab/uncertainty-vst-2026
ER -