Abstract 1099: Leveraging RNA-seq for breast cancer survival analysis and risk stratification

Abstract Advancements in breast cancer treatment have significantly improved patient survival rates; however, the increasing incidence and stratified mortality risks underscore the critical need for Precision Medicine and Artificial Intelligence (AI) tools to provide clinicians with more accurate assessments of patient status. In this study, we analyzed solid tumor RNA-seq data from over 1, 800 breast cancer patients treated at the City of Hope National Medical Center. We employed both supervised and unsupervised machine learning methodologies to develop models aimed at improving the understanding and prediction of patient survival outcomes. Unsupervised clustering of RNA-seq data revealed that, despite conventional clinical groupings (HR+/ER+, HER2+, HR+/HER2+, and TNEG), breast cancer patients may segregate into distinct subgroups with statistically significant differences in survival rates. Notably, HR+/ER+ patients, typically associated with better survival outcomes, were observed to cluster alongside high-risk TNEG patients, challenging conventional survival expectations. Building on these insights, we developed two classes of predictive survival models: (i) binary classifiers to predict survival beyond specific time thresholds (e.g., 3- and 5-year survival) and (ii) probabilistic survival models that estimate survival probabilities for clinician-specified timeframes. Evaluation on a 10% hold-out test set demonstrated strong model performance. The top-performing 3-year survival model achieved AUC’s of 0.88 and 0.84 for the ROC and Precision-Recall curves, respectively, while the 5-year survival model reported elevated AUC’s exceeding 0.95 and 0.98. Furthermore, the probabilistic survival model achieved a robust concordance index (C-index) of 0.83. Our findings, in their current form, may equip clinicians with data-driven solutions to support more accurate diagnoses and better-informed personalized treatment planning. Further integration of these models with EMR data may provide deeper insights into the biological and clinical factors underlying the observed survival disparities among breast cancer patients. These insights could inform both improvements to current clinical practices and the development of novel therapeutic approaches. Citation Format: Pooya Naseri-Nosar, Zahra Eftekhari, Chi Wah Wong. Leveraging RNA-seq for breast cancer survival analysis and risk stratification [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1099.