Abstract 5012: Multi-omics signature for estimating survival time in breast cancer patients

Abstract Introduction: Predicting survival outcomes for breast cancer patients is critical for tailoring personalized treatments. However, effectively integrating multi-omics datasets for survival estimation is a persistent challenge due to the high dimensionality and complexity of the data. To address this, we developed the Multi-Omics Survival Time Estimator (MOST), a novel machine learning framework designed to accurately predict survival time by leveraging multi-omics data. Methods: MOST employs a multi-step pipeline combining stringent feature selection, evolutionary learning, and survival estimation. Using datasets encompassing copy number variations (CNV), microRNA (mRNA), mRNA, and reverse-phase protein array (RPPA) data, the platform identifies a concise yet informative set of biomarkers that contribute to survival time prediction. Feature selection was optimized to ensure relevance and robustness, while evolutionary algorithms enhanced predictive accuracy. The model was rigorously validated to ensure reliability and generalizability. Results: MOST identified a 38-feature multi-omics signature capable of predicting survival time with a correlation coefficient (R) of 0.90 and mean absolute error (MAE) of 1.06 year between predicted and actual survival times. The top 10 features contributing to the prediction included ENSG00000145826.7 and ENSG00000101624.9 from CNVs; hsa-miR-3928 and hsa-miR-326 from miRNAs; KIAA1199, SYT10, PSG3, ACTG1, MYPN from mRNAs; and GUSP4 from RPPA data. These results emphasize the role of integrated genetic, epigenetic, and proteomic markers in defining patient survival. MOST demonstrated superior predictive performance compared to existing methodologies. Conclusions: MOST offers a machine learning framework for integrating multi-omics data to predict survival outcomes in breast cancer patients. Its ability to identify a robust multi-omics signature establishes a foundation for future exploration of multi-omics data across other cancers, reinforcing its clinical relevance and potential translational impact. Citation Format: Srinivasulu Yerukala Sathipati, Manya Reddy Elugoti, Rachel Tillman, Scott Hebbring, Rohit Sharma. Multi-omics signature for estimating survival time in breast cancer patients [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 5012.