Histopathology and proteomics are synergistic for High-Grade Serous Ovarian Cancer platinum response prediction

Oz Kilim,Alex Olar,Andras Biricz,Lilla Madaras,Peter Pollner,Zoltan Szallasi,Zsofia Sztupinszki,Istvan Csabai

DOI: https://doi.org/10.1101/2024.06.01.24308293

2024-06-03

Abstract:Patients with High-Grade Serous Ovarian Cancer (HGSOC) exhibit varied responses to treatment, with 20-30% showing de novo resistance to platinum-based chemotherapy. While hematoxylin-eosin (H&E) pathological slides are used for routine diagnosis of cancer type, they may also contain diagnostically useful information about treatment response. Our study demonstrates that combining H&E-stained Whole Slide Images (WSIs) with proteomic signatures using a multimodal deep learning framework significantly improves the prediction of platinum response in both discovery and validation cohorts. This method outperforms the Homologous Recombination Deficiency (HRD) score in predicting platinum response and overall patient survival. The study sets new performance benchmarks and explores the intersection of histology and proteomics, highlighting phenotypes related to treatment response pathways, including homologous recombination, DNA damage response, nucleotide synthesis, apoptosis, and ER stress. This integrative approach has the potential to improve personalized treatment and provide insights into the therapeutic vulnerabilities of HGSOC.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is: how to improve the prediction accuracy of platinum - based chemotherapy response in patients with High - Grade Serous Ovarian Cancer (HGSOC). Specifically, the paper explores using a multimodal deep - learning framework by combining histopathological images (H&E - stained whole slide images, Whole Slide Images, WSIs) and proteomic features to significantly improve the performance of platinum - based chemotherapy response prediction.

### Background and Motivation
- **Differences in Treatment Response**: Approximately 20% - 30% of HGSOC patients show primary resistance to platinum - based chemotherapy at the time of initial diagnosis, which leads to a poor prognosis for these patients.
- **Limitations of Existing Methods**:
- **H&E - stained Images**: Although H&E - stained images are widely used for routine diagnosis of cancer types, their application in predicting treatment response is not yet sufficient.
- **Proteomics**: Proteomic data can provide information at the molecular level, but lack spatial information.
- **Genomics Methods**: Genomics methods such as HRD score perform well in predicting PARP inhibitor treatment, but have not yet reached the standard for clinical application in predicting platinum - based chemotherapy response.

### Research Objectives
- **Combining Multiple Data Types**: By integrating H&E - stained images and proteomic data, use a multimodal deep - learning model to improve the prediction accuracy of platinum - based chemotherapy response.
- **Surpassing Single - Modal**: Verify whether the multimodal model can surpass the prediction performance of single - modal (such as using only H&E images or only proteomic data).
- **Mechanism Explanation**: Explore the biological mechanisms on which the multimodal model depends when predicting platinum - based chemotherapy response, including pathways such as homologous recombination repair, DNA damage response, nucleotide synthesis, apoptosis, and endoplasmic reticulum stress.

### Main Contributions
- **Performance Improvement**: The research shows that the multimodal model is significantly superior to the single - modal model in predicting platinum - based chemotherapy response.
- **Complementary Information**: The multimodal model can capture complementary information between different modalities, thereby improving the prediction performance.
- **Mechanism Explanation**: Through the attention mechanism and path embedding, the multimodal model can reveal specific histological and proteomic features related to treatment response.

### Conclusion
This study demonstrates the potential of a multimodal deep - learning model that combines histopathological images and proteomic data in predicting platinum - based chemotherapy response in HGSOC patients, providing new tools and methods for personalized treatment.