FunVFPred: Predicting fungal virulence factors using a unified representation learning model

Fungal virulence factors (VFs) are proteins that support host colonization, tissue invasion, immune modulation, and disease progression. Identifying candidate VFs across pathogenic fungi can help characterize mechanisms underlying fungal-host interactions and prioritize proteins for experimental investigation. We developed FunVFPred, a machine learning framework that uses UniRep protein sequence embeddings, and multiple classifiers to predict candidate fungal VFs. Using experimentally validated VFs from human-pathogenic Candida species, the random forest model achieved 73.4% accuracy and a Matthews correlation coefficient (MCC) of 0.47 in 5-fold cross-validation. Evaluation on an independent Aspergillus fumigatus dataset yielded 85.7% accuracy and an MCC of 0.71, demonstrating the potential for cross-species prioritization of virulence-associated proteins. FunVFPred provides an accessible computational resource for investigating fungal pathogenicity and guiding the selection of candidate proteins for downstream experimental validation.