Fungal pathogens cause fatal plant diseases, contributing to up to 20 % loss in annual crop yield. These pathogens accomplish this using an arsenal of secreted effectors that can target the plant apoplast or be translocated into the plant to affect cellular functions and disrupt host defenses. Identification of fungal effectors is a crucial step, not only to advance our understanding of plant–pathogen interactions, but also to assist in the identification of novel disease prevention strategies and ultimately improve food safety and agricultural yields. Given the highly evolving nature of these fungal effectors, there are no recognizable 'signals' that can be used to distinguish fungal effectors from non-effectors. This challenges the few existing tools that have been developed to extract patterns only from the limited number of fungal effectors.
In this work, we present Fungtion (Fungal effector prediction and visualization), a toolkit to accurately predict fungal effectors and visualize their relationship against known fungal effectors. Taking advantage of pre-trained deep learning models, Fungtion learns informative features from the limited fungal effector data with global insights for more robust and accurate prediction. Fungtion additionally provides interactive visualizations to allow users to conduct downstream relationship analyses between predicted and known fungal effectors, which aid in assigning putative functions to these proteins. In summary, by providing enhanced prediction and interactive visualizations, Fungtion is expected to allow biologists to characterize putative fungal effectors and design new hypotheses.