CircExor enables interpretable prediction of circRNA localization into extracellular vesicles.

CircExor predicts circRNA localization into extracellular vesicles with an AUROC of 0.743, outperforming existing tools and aiding biomarker discovery.

• Why it matters: Understanding how circRNAs are sorted into EVs is crucial for developing biomarkers and understanding intercellular communication, yet current models are inadequate for circRNA-specific prediction.
• What they did: They developed circExor, a machine learning framework using a curated dataset of 2102 circRNAs, incorporating sequence encoding strategies tailored for circular topology and length variability.
• The result: CircExor offers accurate, interpretable predictions linking sequence features to EV localization, identifying candidate RBPs like YBX1 and HNRNPK, thus facilitating mechanistic insights and downstream biomarker research.