Predicting Mutational Signature Exposures from H&E Whole Slide Images: A Pan-Cancer Feasibility Study
Abstract
Mutational signatures reveal cancer-driving processes with clinical relevance: MMR-deficient tumors respond better to immunotherapy, HRD tumors are sensitive to PARP inhibitors, and POLE-mutant tumors often have high mutation burdens influencing treatment response. Yet signature profiling remains limited by the cost, complexity, and turnaround time of whole-genome sequencing (WGS) or whole-exome sequencing. Histopathology is widely available and cost-effective, and H&E morphology has been linked to MSI, HRD, POLE-related processes, and driver mutations. Whether histology can recover the broader landscape of mutational signature exposures in a pan-cancer setting remains unknown. Here we introduce Hist2Sig, a deep learning framework predicting exposures to 30 COSMIC SBS signatures from H&E-stained slides. Trained on matched WGS and histology data from 7,063 TCGA patients across 29 cancer types, Hist2Sig was compared with a tumor type-only baseline to separate morphology-derived signal from tissue-of-origin priors. In internal cross-validation, it recovered relative signature compositions across most tumor types and achieved higher mean top-three overlap than the baseline in 19 of 29, with the largest gains in COAD, PRAD, GBM, and UCEC. It also recurrently recovered signatures of poorly characterized etiology, including SBS8, SBS12, SBS39, and SBS40a. In an external CPTAC cohort of 193 patients across five tumor types, Hist2Sig retained moderate concordance with observed compositions and outperformed the baseline in glioblastoma and pancreatic adenocarcinoma. These results support the feasibility of inferring mutational signature exposures from routine histology, while highlighting variation across tumor types. Hist2Sig provides a foundation for tumor-specific models that could complement sequencing in pre-sequencing triage or support decisions where WGS is unavailable.
Related
- Method: Selection of anatomical sites for predicting in vivo body composition in pig and goat using tissue thicknesses measured from computed tomography images
- Correlation-aware discovery of co-occurring mutational signatures in cancer
- Human macrophages encode stimulus-specific information of prior exposures through trained immunity
- Deep mutational scanning of the GPCR rhodopsin reveals pleiotropic mutational effects and their roles in modulating receptor signaling
- Chlamydia trachomatis Incidence in Relation to Vaginal Microbiota Dynamics, Immunogenetics and Exposures in a Cohort of Young Student Women in France
BibTeX
@misc{sartori2026predicting,
title = {{Predicting Mutational Signature Exposures from H&E Whole Slide Images: A Pan-Cancer Feasibility Study}},
author = {Flavio Sartori and Cesare Rollo and Isabella Caranzano and Tiziana Sanavia and Piero Fariselli and Corrado Pancotti},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.30985},
eprint = {2609.30985},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.30985},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Predicting Mutational Signature Exposures from H&E Whole Slide Images: A Pan-Cancer Feasibility Study
AU - Flavio Sartori
AU - Cesare Rollo
AU - Isabella Caranzano
AU - Tiziana Sanavia
AU - Piero Fariselli
AU - Corrado Pancotti
PY - 2026
DA - 2026/09/25
PB - arXiv
DO - 10.48550/arxiv.2609.30985
UR - https://arxiv.org/abs/2609.30985
AB - Mutational signatures reveal cancer-driving processes with clinical relevance: MMR-deficient tumors respond better to immunotherapy, HRD tumors are sensitive to PARP inhibitors, and POLE-mutant tumors often have high mutation burdens influencing treatment response. Yet signature profiling remains limited by the cost, complexity, and turnaround time of whole-genome sequencing (WGS) or whole-exome sequencing. Histopathology is widely available and cost-effective, and H&E morphology has been linked to MSI, HRD, POLE-related processes, and driver mutations. Whether histology can recover the broader landscape of mutational signature exposures in a pan-cancer setting remains unknown. Here we introduce Hist2Sig, a deep learning framework predicting exposures to 30 COSMIC SBS signatures from H&E-stained slides. Trained on matched WGS and histology data from 7,063 TCGA patients across 29 cancer types, Hist2Sig was compared with a tumor type-only baseline to separate morphology-derived signal from tissue-of-origin priors. In internal cross-validation, it recovered relative signature compositions across most tumor types and achieved higher mean top-three overlap than the baseline in 19 of 29, with the largest gains in COAD, PRAD, GBM, and UCEC. It also recurrently recovered signatures of poorly characterized etiology, including SBS8, SBS12, SBS39, and SBS40a. In an external CPTAC cohort of 193 patients across five tumor types, Hist2Sig retained moderate concordance with observed compositions and outperformed the baseline in glioblastoma and pancreatic adenocarcinoma. These results support the feasibility of inferring mutational signature exposures from routine histology, while highlighting variation across tumor types. Hist2Sig provides a foundation for tumor-specific models that could complement sequencing in pre-sequencing triage or support decisions where WGS is unavailable.
ER -