Extracellular vesicles (EVs) are membrane-bound vesicles that are released by all cells into the extracellular environment. They play crucial roles in intercellular communication and the regulation of physiological and pathological processes. These vesicles are composed of a lipid bilayer and contain diverse cargo such as proteins, nucleic acids, and metabolites. The molecular composition of EV cargo mirrors the physiological condition of the parent cell and can profoundly influence recipient cells by modulating gene expression, signalling pathways, and functional attributes. Research into EVs has expanded significantly, identifying them as promising diagnostic biomarkers for various diseases and potential tools for therapeutic applications. However, identifying the source of EVs from complex biological samples such as blood, urine, and plasma remains challenging due to their heterogeneous cellular origins.
we introduce EV Barcode, an open-source web resource leveraging machine learning techniques to predict the cellular source of EVs based on their cargo composition. The predictive model was trained on extensive data comprising over 5 million single-cell data from more than 1000 samples representing diverse tissues/sites. Model refinement involved hyperparameter optimization, feature selection, and ensemble techniques, supported by rigorous cross-validation to ensure robust performance. EV Barcode represents a significant advancement in EV research, offering EV researchers worldwide a powerful tool to predict the cellular origin of EVs identified in omics experiments. By filling critical knowledge gaps in the field, this resource promises to enhance our understanding of EV biology and accelerate the translation of EV research into clinical applications.