Breast cancer has become the leading cause of global cancer incidence and a serious threat to women’s health. Accurate diagnosis and early treatment are of great importance to prognosis. Although clinically used diagnostic approaches can be used for cancer screening, accurate diagnosis of breast cancer is still a critical unmet need. Here, we report a 4-plex droplet digital PCR technology for simultaneous detection of four small extracellular vesicle (sEV)-derived mRNAs (PGR, ESR1, ERBB2 and GAPDH) in combination with machine learning (ML) algorithms to improve breast cancer diagnosis. We evaluate the diagnsotic results with and without the assistance of the ML models. The results indicate that ML-assisted analysis exhibits higher diagnostic performance even using a single marker for breast cancer diagnosis, and demonstrate improved diagnostic performance under the best combination of biomarkers and suitable ML diagnostic model. Therefore, multiple sEV-derived mRNAs analysis coupled with ML not only provides the best combination of markers for breast cancer diagnosis, but also significantly improves the diagnostic efficiency of breast cancer.

乳腺癌已成为全球癌症发病率的首要原因，严重威胁女性健康。准确诊断和早期治疗对预后非常重要。尽管临床上使用的诊断方法可用于癌症筛查，但准确诊断乳腺癌仍然是一个尚未得到满足的关键需求。在这里，我们报告了一种 4 重液滴数字 PCR 技术，可结合机器学习 (ML) 算法同时检测四种小细胞外囊泡 (sEV) 衍生的 mRNA（PGR、ESR1、ERBB2 和 GAPDH），以改善乳腺癌诊断。我们在有和没有 ML 模型的帮助下评估诊断结果。结果表明，即使使用单一标记物进行乳腺癌诊断，ML 辅助分析也表现出更高的诊断性能，并证明在生物标志物和合适的 ML 诊断模型的最佳组合下提高了诊断性能。因此，多个sEV衍生的mRNAs分析结合ML不仅为乳腺癌诊断提供了最佳标志物组合，而且显着提高了乳腺癌的诊断效率。