Abstract LB409: OncodeAi™-Breast CED: A nanoparticle-enhanced IR molecular sensing platform for early breast cancer detection

Abstract Current methodologies for early breast cancer screening, such as mammography, often fail to detect tumors at their earliest stages, particularly in dense breast tissue. OncodeAi™-Breast Cancer Early Detection (CED) platform addresses this challenge by utilizing plasmonic nanoparticle- and chemical-enhanced infrared (IR) molecular sensing combined with advanced machine learning (ML) to identify distinct molecular fingerprints in patient blood samples. Objective: In this study, we aimed to evaluate the performance of OncodeAi™-Breast CED in detecting early-stage breast cancer using blinded blood samples prospectively collected from patients diagnosed with breast cancer and age-matched healthy controls. Methods: For model development, serum samples were collected from retrospective cohorts (N=4,120) consisting of healthy controls (n=2,263), breast cancer patients from stages 0 to IV (n=752), and a mixture of various cancer types (n=1,105) and used to train an ML model to identify molecular fingerprints indicative of breast cancer. Fourier-transform IR spectroscopy was employed to obtain the spectral molecular fingerprint for each sample, with the methodology enhanced by resonant surface-enhanced IR absorption (rSEIRA) facilitated by a proprietary nano-structured IR reflective sampling substrate. A deep learning-based ML algorithm was trained using the captured molecular fingerprint data based on an 80/20 training/validation split. For testing, an optimized model was deployed to identify the unknown samples from a blinded, prospectively collected serum samples from an independent multi-center cohort (N=185) that included breast cancer patients at various stages (n=60, Stages 0-III) and healthy controls (n=125). Blinded predictions were performed on the prospective samples across seven independent, twice-randomized technical repeats. Results: Our rSEIRA/ML-based OncodeAi™ platform achieved an average of 96.4% sensitivity (95% CI: 93.6%-99.3%) and 96.9% specificity (95% CI: 95.1%-98.7%) in detecting breast cancer of all stages across multiple technical repeats, with an average ROC AUC of 0.99. The average sensitivity for detecting early-stage disease (Stages 0-I; n=49) was 98.3% (95% CI: 96.0%-100%), and 10 out of 11 breast cancer patients with either stage II (n=6), III (n=3), or unknown stage (n=2) were correctly identified. Conclusion: OncodeAi™-Breast CED marks a transformation in breast cancer detection as an early disease screening tool. By integrating the precision of IR molecular fingerprinting with the advanced capabilities of nanotechnology and machine learning, OncodeAi™ delivers unparalleled accuracy in detecting early-stage cancer. Its non-invasive nature, low cost, high scalability, and superior sensitivity offer transformative potential in population-wide screening programs, especially for high-risk and underserved communities. This platform also holds potential for expansion to other malignancies, including lung and colorectal cancer. Citation Format: Ali Khammanivong, Eng Hock Lee, Darlyna Khonkhammy, Caden Gunnarson, Ganesh D. Sockalingum, Saad Bedros, Robert Petcavich, Dan Que Pham. OncodeAi™-Breast CED: A nanoparticle-enhanced IR molecular sensing platform for early breast cancer detection [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_2):Abstract nr LB409.