Abstract 1843: Integrating machine learning with microfluidic technologies for proteomic profiling of extracellular vesicles in triple-negative breast cancer

Abstract

Abstract Accurate diagnosis of breast cancer using circulating biomarkers present in plasma remains an important challenge. In particular, protein changes in tumor-derived extracellular vesicles (tdEVs) have emerged as potential biomarkers for breast cancer diagnosis because they accurately reflect dynamic changes in tumors. In this study, we compared the proteomes of extracellular vesicle (EV) isolated from the plasma of 100 breast cancer patients and 30 healthy individuals who visited Severance Hospital between Mar 2010 and Dec 2015. Microfluidic chip-based protocol that facilitates the removal of contaminants such as albumin and immunoglobulins was used for the extraction of enriched tdEVs with small amounts of plasma. Comparative analysis of the proteomes identified 26 significant biomarkers that could indicate differences between breast cancer patients and healthy individuals. Using the LsBoost-CNN-SVM hybrid machine learning algorithm, we especially identified key EV protein biomarkers for detecting triple-negative breast cancer (TNBC), a breast cancer subtype with a poor prognosis and a high recurrence rate, as well as lacking therapeutic and diagnostic targets. Remarkably, a signature consisting of three EV proteins, specifically extracellular matrix protein 1 (ECM1), mannose-binding lectin 2 (MBL2), and biotinidase (BTD), effectively distinguished TNBC patients from healthy individuals. In our proteomic analysis set (n=73), this signature exhibited impressive performance, achieving a sensitivity of 93.3% and specificity of 93% in the accurate discrimination of TNBC from the control group. The validation set (n=40) confirmed these findings, with 100% sensitivity and 80% specificity. This signature not only served as a diagnostic tool but also provided valuable insights into the risk of recurrence and patient prognosis. We found a novel diagnostic approach that holds the potential to revolutionize breast cancer diagnostics by enhancing the reliability of tumor-related information obtained from blood samples. Citation Format: Min Woo Kim, Jee Ye Kim, Young Kim, Suji Lee, Sol Moon, Ju-yong Hyon, Kyung-A Hyun, Yeji Yang, Seongmin Ha, Sunyoung Park, Hogyeong Gawk, Haeji Lee, Eun Hee Han, Jin Young Kim, Hyo-Il Jung, Young-Ho Chung, Seung Il Kim. Integrating machine learning with microfluidic technologies for proteomic profiling of extracellular vesicles in triple-negative breast cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 1843.