ABSTRACT
BACKGROUND: According to the global cancer burden data released in 2020, breast cancer (BC) has become the most common cancer in the world. Similar to those of other cancers, the present methods used in clinic for diagnosing early BC are invasive, inaccurate, and insensitive. Hence, new non-invasive methods capable of early diagnosis are needed. METHODS: We applied next-generation sequencing and analyzed the messenger RNA (mRNA) profiles of plasma extracellular vesicles (EVs) derived from 14 BC patients and 6 patients with benign breast lesions. We used 3 regression models, namely support vector machine (SVM), linear discriminate analysis (LDA), and logistic regression (LR), to develop classifiers for use in making predictive BC diagnoses; and used 259 plasma samples, including those obtained from 144 patients with BC, 72 patients with benign breast lesions, and 43 healthy women, which were divided into training groups and validation groups to verify their performances as classifiers by quantitative reverse transcription polymerase chain reaction (RT-qPCR). The area under the curve (AUC) and accuracy, sensitivity, and specificity of the classifiers were cross-validated with the leave-1-out cross-validation (LOOCV) method. RESULTS: Among all combinations assessed with the 3 different regression models, an 8-mRNA combination, named EXOBmRNA, exhibited high performance [accuracy =71.9% and AUC =0.718, 95% confidence interval (CI): 0.652 to 0.784] in the training cohort after LOOCV was performed, showing the largest AUC in the SVM model. The mRNAs in EXOBmRNA were HLA-DRB1, HAVCR1, ENPEP, TIMP1, CD36, MARCKS, DAB2, and CXCL14. In the validation cohort, the AUC of EXOBmRNA was 0.737 (95% CI: 0.636 to 0.837). In addition, gene function and pathway analyses revealed that different levels of gene expression were associated with cancer. CONCLUSIONS: We developed a high-performing predictive classifiers including 8 mRNAs from plasma extracellular vesicles for diagnosing breast cancer.