Machine-Learning Prediction of Extracellular Vesicle Protein Sorting Expands the Characterization of Secretory Functions in Mucor circinelloides

Opportunistic infections are a growing concern in medical mycology. In that context, Mucor circinelloides is recognized as an etiological agent of emerging life-threatening infections in humans. Furthermore, this biomass-degrading species is important in the biotechnological context due to its potential for enzyme discovery. In this work, we describe an in silico pipeline for the comprehensive characterization of the secretome of this species. Our analyses suggest that M. circinelloides encodes a diverse enzymatic repertoire, including multiple carbohydrate-active enzymes (CAZymes). Functional characterization of this secreted protein set identified candidate proteins with potential relevance for industrial application and virulence-related processes. Importantly, we extended this framework by integrating a sequence-based machine-learning (ML) approach for the prediction of extracellular vesicle (EV)-associated proteins. Using a Random Forest model trained on high-confidence EV datasets, we predicted a subset of candidate proteins in M. circinelloides and explored their functional distribution. The predicted EV proteome displayed a degree of similarity to experimentally characterized human EV proteomes, suggesting that the model is able to capture sequence-derived predictive signals. Overall, our results support the potential of ML as a scalable strategy to be integrated into the study of fungal EVs. This approach provided a foundation for research aimed at experimentally validating EV cargo and refining predictive models. Beyond M. circinelloides, we provide a framework that may be useful for future studies on EV biology across diverse fungi of medical and biotechnological interest.