Towards Trustworthy Biological Alignment in TabPFN-Probed Pathology Foundation Models
Abstract
Histology and transcriptomics provide complementary views of tissue biology, capturing spatial morphology and molecular activity, respectively. Pathology foundation models (PFMs) learn rich morphological representations from H&E images, yet strong downstream performance alone does not establish whether these representations encode biologically meaningful and robust molecular information. We present a **training-free framework for auditing biological alignment in frozen PFMs** using spatially paired histology and transcriptomics from HEST-1k, evaluated on **240 samples spanning three organs**. Multiple frozen PFMs are used to extract H&E representations, while gene expression is aggregated into biologically interpretable pathway-level programs. We use TabPFN as a pretrained probe to quantify the extent to which these molecular programs can be decoded from frozen image representations without task-specific gradient updates. Beyond predictive performance, our audit examines whether pathway decodability generalizes across tissue sections, patient groups, and tissue types; whether representations exhibit section-level or other shortcut dependencies; and whether predictions remain stable under small image perturbations and context resampling. This multi-tissue evaluation distinguishes molecular programs that are consistently encoded from those that are tissue-specific, unstable, or shortcut-sensitive. Our framework therefore provides a systematic approach for assessing not only **what biological information pathology foundation models encode, but also how reliably that information persists under clinically relevant sources of variation**.
Related
- Trustworthy ML/AI for Aging Clocks: Preventing Systematic Prediction Bias in Biological Age Estimation
- Alignment with sex-typical brain architecture: Links to mood pathology, sex hormones and transcriptomics of neurodegenerative conditions
- Scaling subjects in cross-modal alignment: video decoding with EEG foundation model
- WILSON - Multi-magnification composites as a foundation for patient-level pathology analysis and diagnostic text generation
- Decomposing scanner batch effects and their impact on pathology foundation models
BibTeX
@misc{bhattacharjee2026towards,
title = {{Towards Trustworthy Biological Alignment in TabPFN-Probed Pathology Foundation Models}},
author = {Ushashi Bhattacharjee and Alloy Das and Saria Hannan and Tirtho Roy and Koushik Howlader and Soumik Sarkar},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.29523},
eprint = {2609.29523},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.29523},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Towards Trustworthy Biological Alignment in TabPFN-Probed Pathology Foundation Models
AU - Ushashi Bhattacharjee
AU - Alloy Das
AU - Saria Hannan
AU - Tirtho Roy
AU - Koushik Howlader
AU - Soumik Sarkar
PY - 2026
DA - 2026/09/01
PB - arXiv
DO - 10.48550/arxiv.2609.29523
UR - https://arxiv.org/abs/2609.29523
AB - Histology and transcriptomics provide complementary views of tissue biology, capturing spatial morphology and molecular activity, respectively. Pathology foundation models (PFMs) learn rich morphological representations from H&E images, yet strong downstream performance alone does not establish whether these representations encode biologically meaningful and robust molecular information. We present a **training-free framework for auditing biological alignment in frozen PFMs** using spatially paired histology and transcriptomics from HEST-1k, evaluated on **240 samples spanning three organs**. Multiple frozen PFMs are used to extract H&E representations, while gene expression is aggregated into biologically interpretable pathway-level programs. We use TabPFN as a pretrained probe to quantify the extent to which these molecular programs can be decoded from frozen image representations without task-specific gradient updates. Beyond predictive performance, our audit examines whether pathway decodability generalizes across tissue sections, patient groups, and tissue types; whether representations exhibit section-level or other shortcut dependencies; and whether predictions remain stable under small image perturbations and context resampling. This multi-tissue evaluation distinguishes molecular programs that are consistently encoded from those that are tissue-specific, unstable, or shortcut-sensitive. Our framework therefore provides a systematic approach for assessing not only **what biological information pathology foundation models encode, but also how reliably that information persists under clinically relevant sources of variation**.
ER -