Liquid biopsy is a promising non-invasive approach for early detection of cancer and monitoring disease progression and therapeutic response. We have recently developed a powerful and reliable method to isolate circulating small extracellular vesicles (sEVs) from plasma of breast cancer patients and showed that semi-quantitative proteomic profiling of sEVs can be used for early detection of breast cancer. Our preliminary analysis also suggests that the sEVs proteome could be used for prediction of recurrent disease. In this proof-of-concept project, we proposed not only to improve our isolation and profiling methods of circulating sEVs, but also to identify possible connections between the transcriptome of the primary breast tumors and the proteome of the circulating sEVs. To this end, we plan to profile matching samples of sEVs and primary tumors from breast cancer patients and to establish prediction models by machine learning as well as connectivity map between the tumor transcriptome and the sEVs proteome using computational methods. Deconvolution models will be applied to stratify sEVs based on their cells of origin. This combined analysis may predict sEVs biomarkers for distinct properties of the primary tumors, including tumor stage, tumor composition, heterogeneity, and response to treatment.