Abstract PS3-06-24: Transcriptomics-guided AI outperforms and expands on OncotypeDX to predict Hormonal therapy response and Chemotherapy benefit from H&E-stained Whole Slide images in ER+/HER2- Breast Cancer

Abstract Purpose Oncotype DX (ODX) is one of the most widely used genomic tests for risk stratification and prediction of response to hormonal therapy (HT) of early stage ER+/HER2- node-negative breast cancer. However, this test does not provide guidance on the extent of chemotherapy (CT) benefit. We developed an AI-based model that predicts response to HT and failure of HT+CT directly from H 0.002). Fifteen patients with ODX20 treated with HT alone metastasized and our AI model identified 11/15 (73%) of these patients as HT-failures. The 150 patients treated with HT+CT based on ODX scores ranging from 26 to 71 were stratified using our CT-failure biomarker with an HR of 2.42 (p 0.02), identifying 22/35 (65%) patients who metastasized after being treated with HT+CT. In a multivariate analysis, ODX score added prognostic information to the AI- based HT-response biomarker (p=0.04) and age added prognostic information to the AI- based HT+CT-failure biomarker (p=0.03). By integrating signaling pathway alterations together with changes in gene expressions, we were able to utilize these enriched genomic changes using our AI model to identify responders and non-responders to HT and HT+CT. Conclusion To our knowledge, this is the first AI model that can predict response to HT and HT+CT in early stage ER+/HER2- node-negative breast cancer. Our AI model by itself identified 73% patients with low ODX scores who developed metastasis following HT alone and 63% patients with high ODX who did not benefit from HT+CT. Combining ODX recurrence scores with the output of our AI model improved the prediction of response in HT treated patients. Citation Format: H. Muhammad, S. S. Chavan, C. Feng, D. K. Almaraz, H. Basu, W. Huang, R. Roy, G. Wilding, G. Mills, S. Kummar, S. Krishnamurthy. Transcriptomics-guided AI outperforms and expands on OncotypeDX to predict Hormonal therapy response and Chemotherapy benefit from H 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS3-06-24.