What question did this study set out to answer?

This research aims to predict genomic alterations from H&E whole-slide images to improve molecular testing strategies.

April 10, 2026Open Access

Pan-Cancer Prediction of Genomic Alterations from H&E Whole-Slide Images in a Real-World Clinical Cohort

Key Points

This research aims to predict genomic alterations from H&E whole-slide images to improve molecular testing strategies.
Retrospective enrollment of 437 patients across 26 cancers with clinical gene-panel data
Extraction of binary endpoints for genomic alterations across various categories
Utilization of five pathology foundation models for slide embedding extraction and benchmarking
Training a multi-instance learning model based on the best-performing patch feature system
Titan model achieved a median Area Under the Receiver Operating Characteristic curve (AUROC) of 0.77 for slide benchmarking
At the patch-level, predicting APC_SNV reached an AUROC of 0.916
KRAS_SNV prediction obtained an AUROC of 0.811 in the test set

Abstract

Background: Predicting genomic alterations from routine hematoxylin and eosin (H at the patch-level, prediction of APCSNV reached an AUROC of 0. 916, and prediction of KRASSNV reached an AUROC of 0. 811 on the held-out test set. Conclusions: In a heterogeneous clinical gene-panel setting, pathology foundation models can provide strong baseline genomic-prediction signals without additional fine-tuning. We propose a practical, deployment-oriented two-stage workflow: rapid slide-embedding screening to prioritize robust representations and candidate endpoints, followed by patch-level training for high-value tasks where additional performance gains and interpretable regions are clinically worthwhile.

Connected Papers

Building similarity graph...

Analyzing shared references across papers

Discussion

Cite this study

Ma et al. (Wed,) studied this question.

synapsesocial.com/papers/69d893406c1944d70ce04486 — DOI: https://doi.org/10.3390/genes17040371

Authors

Dongheng Ma

Osaka City University

Hinano Nishikubo

Osaka City University

Tomoya Sano

Kurume University

Journals

Genes

Actions

Institutions

Osaka City University

References and Citations

Connected Papers

Building similarity graph...

Analyzing shared references across papers

Pan-Cancer Prediction of Genomic Alterations from H&E Whole-Slide Images in a Real-World Clinical Cohort

Key Points

Abstract

Citation Network

Connected Papers

Discussion

Cite this study

Authors

Journals

Actions

Institutions

References and Citations

Citation Network

Connected Papers

Discussion