Independent benchmark of H&E-based gene expression prediction in skin

Abstract

Given the widespread availability of H&E slides, there is considerable interest in determining whether molecular information can be inferred directly from tissue morphology, potentially reducing the need for costly spatial transcriptomic profiling. We assessed three state-of-the-art methods for predicting single-cell gene expression from H&E images across three skin disease contexts and two Xenium panels. As controls, we included simple linear regression models trained on embeddings from multiple foundation models, totalling 16 models evaluated in this study. We show that all models performed poorly: for most genes, prediction accuracy was near zero, and reliable predictions were largely restricted to keratinocyte-associated genes. Predicted expression failed to preserve cell-type identity and spatial organisation, with only keratinocytes forming coherent clusters, while immune, fibroblast, and other dermal populations were extensively mixed. Notably, simple ridge regression on pretrained embeddings matched or outperformed the more complex published architectures, indicating that the predictive signal originates primarily from image representations rather than model design. Our results demonstrate that current H&E-based gene expression prediction methods are not yet suitable for single-cell-level interpretation of spatial transcriptomics in skin tissue.

Octave: Scale-resolved Evaluation of Spatial Gene Expression Prediction from Histology
Q. Wang, S. Gu, Q. Lan +3
bioRxiv25 Sep 2026PreprintAll rights reservedFull textBioinformatics & computational biologybioRxiv
Syntaxin11 deficiency inhibits CRAC channel priming to suppress cytotoxicity and gene expression in T lymphocytes independent of membrane trafficking.
Sritama Datta, Abhikarsh Gupta, Kunal M Jagetiya +14
ELife19 Oct 2026Peer-reviewedCC BYFull textNeuroscienceEurope PMC
Disease relevance and replicability of deep learning gene expression prediction
A. Zhang, S. Tasaki, D. Connell +2
bioRxiv24 Sep 2026PreprintCC BYFull textGenetics & genomicsbioRxiv
Alpha Spatial: Language-guided prediction of single-cell spatial gene expression from histology
X. YANG, Z. GUO, X. GUO +6
bioRxiv8 Oct 2026PreprintCC BYFull textBioinformatics & computational biologybioRxiv
Mendel, a foundation model of human genetic variation, prioritizes regulatory variants and improves gene expression prediction
A. Salman, H. Feng, P. Wei +4
bioRxiv28 Sep 2026PreprintCC BYFull textGenetics & genomicsbioRxiv

Cite

BibTeX RIS

BibTeX

@misc{shaikhutdinova2026independent,
  title = {{Independent benchmark of H&E-based gene expression prediction in skin}},
  author = {R. Shaikhutdinova and S. Gansberger and J. Staller and N. Singh and I. Oyarzun and M. Simon and B. Sterniczky and P. Tschandl and J. Griss},
  howpublished = {bioRxiv},
  year = {2026},
  month = sep,
  doi = {10.64898/2026.09.07.749926},
  url = {https://www.biorxiv.org/content/10.64898/2026.09.07.749926v1},
  note = {CC BY-NC-ND},
}

RIS

TY  - UNPB
TI  - Independent benchmark of H&E-based gene expression prediction in skin
AU  - R. Shaikhutdinova
AU  - S. Gansberger
AU  - J. Staller
AU  - N. Singh
AU  - I. Oyarzun
AU  - M. Simon
AU  - B. Sterniczky
AU  - P. Tschandl
AU  - J. Griss
PY  - 2026
DA  - 2026/09/07
PB  - bioRxiv
DO  - 10.64898/2026.09.07.749926
UR  - https://www.biorxiv.org/content/10.64898/2026.09.07.749926v1
AB  - Given the widespread availability of H&E slides, there is considerable interest in determining whether molecular information can be inferred directly from tissue morphology, potentially reducing the need for costly spatial transcriptomic profiling. We assessed three state-of-the-art methods for predicting single-cell gene expression from H&E images across three skin disease contexts and two Xenium panels. As controls, we included simple linear regression models trained on embeddings from multiple foundation models, totalling 16 models evaluated in this study. We show that all models performed poorly: for most genes, prediction accuracy was near zero, and reliable predictions were largely restricted to keratinocyte-associated genes. Predicted expression failed to preserve cell-type identity and spatial organisation, with only keratinocytes forming coherent clusters, while immune, fibroblast, and other dermal populations were extensively mixed. Notably, simple ridge regression on pretrained embeddings matched or outperformed the more complex published architectures, indicating that the predictive signal originates primarily from image representations rather than model design. Our results demonstrate that current H&E-based gene expression prediction methods are not yet suitable for single-cell-level interpretation of spatial transcriptomics in skin tissue.
ER  -

Abstract

Related

Cite