Abstract PS1-06-12: Early-stage breast cancer detection with a plasma cell-free RNA and AI-based liquid biopsy platform

Abstract Background. Early detection of breast cancer saves lives, yet mammography—the current standard for screening—has well-documented limitations. Its sensitivity is notably reduced in women with dense breast tissue, where tumor tissue may be obscured by the naturally dense parenchyma, leading to false-negative findings. Emerging blood-based biomarkers for early cancer detection and monitoring have the potential to augment existing image-based screening technologies in high-risk populations including dense breasts. Tumor-derived RNA has two key advantages compared to tumor-derived DNA which favor its application in diagnosis of small tumors; i) each cell can produce multiple copies of RNA, while DNA content is limited to cellular ploidy, and ii) while cfDNA can originate from any genomic region, cfRNA can only originate from the subset of accessible chromatin undergoing active transcription. We have previously identified and developed a novel category of cancer-associated, small orphan non-coding RNAs (oncRNAs), and combined them with generative AI modeling to develop a liquid biopsy platform requiring a small volume of plasma. This platform has demonstrated high sensitivity and specificity for detection of early stage disease across several cancer types. Here we developed a cell-free RNA and AI-based assay to detect breast cancer across the range of tumor stages and sizes and analyzed it in a separate, independent test set. Methods. We utilized The Cancer Genome Atlas (TCGA) small RNA profiles to discover a library of pan-cancer oncRNAs that were significantly enriched among breast tumors compared to adjacent normal tissues. Our plasma-based training cohort comprised of 745 female patients with clinically diagnosed, pathologically confirmed untreated breast cancer and 258 age-matched women with no known history of cancer from five different sources (mean age 59.3 years ± 13.4 S.D.), including 593 stage I/II (79.6%), 147 stage III/IV (19.7%), and 5 unknown stage (0.7%) patients. The test set was obtained from a different sample source and consisted of blood samples from 92 female patients with breast cancer and 98 individuals with no known diagnosis of cancer (mean age 60 years ± 10.8 S.D.). The test-set included 60 stage I/II (65.22%) and 32 stage III/IV (34.78%) patients. We processed 1 mL of plasma for each sample through our universal, automated cell-free RNA workflow and sequenced at an average depth of 53 million ± 20 S.D. 100 bp single-end reads. We trained a generative AI model using 5-fold cross-validation (CV) to predict presence or absence of invasive cancer and then applied it on the independent test set. Results. Our model achieved an overall AUC of 0.88 (95% CI: 0.85-0.9) for predicting breast cancer versus controls in CV, and an overall AUC of 0.89 (0.83-0.93) in the independent test set. At 90% specificity, the model achieved overall cross-validated sensitivity of 71.8% (63.7%-77.6%) in the training data, and sensitivity of 71.7% (37.4%-79.5%) in the test set. For stage I breast cancer, the model had a CV sensitivity of 63% (57%-68%) in the training data, and 63% (44%-80%) in the test set, both at a 90% specificity. Conclusions. In an independent cohort, the oncRNA and AI-based plasma assay demonstrated high sensitivity for detecting early-stage invasive breast cancer. Given that 40-50% of women have dense breast tissue—where traditional imaging methods are less effective—and assay sensitivity of 70% in an independent cohort, our findings support the potential of oncRNA-based liquid biopsies for screening high-risk populations after negative mammography. A non-invasive blood test could complement current standard screening methodologies in women with dense breasts by stratifying women needing supplemental screening. A prospective study is planned to further validate these findings. Citation Format: L. Schwartzberg, M. Karimzadeh, A. Momen-Roknabadi, T. B. Cavazos, N. Chen, J. Wang, M. Multhaup, J. Ku, A. Krishnan, M. Hernandez, R. Hanna, L. Fish, M. Gebala, H. Goodarzi, H. Li, F. Hormozdiari, B. Behsaz, A. Hartwig, B. Alipanahi. Early-stage breast cancer detection with a plasma cell-free RNA and AI-based liquid biopsy platform abstract. In: Proceedings of the San Antonio Breast Cancer Symposium 2025; 2025 Dec 9-12; San Antonio, TX. Philadelphia (PA): AACR; Clin Cancer Res 2026;32(4 Suppl):Abstract nr PS1-06-12.