In a recent preprint published on bioRxiv, researchers from Memorial Sloan Kettering Cancer Center, the Technical University of Dresden, the European Institute of Oncology, MultiplexDX and other institutions report on the development of multimodal histopathologic models that can stratify hormone receptor-positive early breast cancer.

For patients with early, hormone receptor-positive breast cancer (without HER2 amplification), genetic tests like Oncotype DX can help identify those who would benefit most from chemotherapy in addition to hormone therapy. However, these genetic tests can be expensive and take time, limiting their widespread use. This study investigated whether information from standard pathology slides (hematoxylin and eosin or H&E stained) and associated text reports could be used instead to predict the Oncotype DX recurrence score, which indicates a patient's risk of cancer recurrence. The researchers developed a machine learning model called Orpheus that was trained on data from over 6,000 patients. Orpheus was able to accurately predict the Oncotype DX recurrence score from just the H&E slides, the text reports, or a combination of the two. The model performed well in predicting which patients had a high-risk recurrence score, doing better than existing methods based on clinical and pathological features alone. Importantly, the model also worked well when tested on independent patient datasets from other institutions.

The study utilized data from MultiplexDX, a company that provides innovative solutions for breast cancer diagnosis and therapy response prognosis. The availability of high-quality data from MultiplexDX was crucial for the success of this project. By integrating different data types, they were able to develop models that provide much more accurate risk stratification compared to traditional clinical and pathological parameters alone. This is an exciting example of how advanced computational approaches, when applied to comprehensive, multimodal data, can unlock new insights in cancer biology and clinical management.

The preprint is available on the bioRxiv preprint server, and the researchers are currently preparing the manuscript for peer-reviewed publication. The findings have the potential to impact the clinical care of patients with hormone receptor-positive early breast cancer.

Find our more HERE https://www.biorxiv.org/content/10.1101/2024.02.23.581806v1