Artificial intelligence-based phenotypic screening using Aspergillus oryzae identified a novel antifungal RK-1205-I

When treated with various compounds, fungi exhibit characteristic morphological changes depending on their mode of action. Previously, we constructed morphology-based databases of the rice blast fungus Pyricularia oryzae and human pathogenic fungus Candida albicans and used them for antifungal screening. As these databases are manually created by human experts, objectivity and throughput may be a concern. To overcome these limitations, we developed a new artificial intelligence (AI)-based automated classification system for morphological changes in the filamentous fungus Aspergillus oryzae. Using this system, we screened a library of 7602 microbial broths and found that the culture broth of Streptomyces sp. RK21-A1205 exhibited potent antifungal activity and induced unique morphological changes distinct from those induced by conventional antifungal agents. Antifungal-activity-guided purification yielded the active metabolite RK-1205-I (1), whose chemical structure was identified as a new fostriecin derivative. 1 showed strong growth inhibitory activity against A. oryzae and other fungi, including A. fumigatus and C. auris, with IC₅₀ value of 0.020 and 0.21 µM, respectively. Furthermore, 1 inhibited protein phosphatase 2A (PP2A) activity in the fungal cells. This study demonstrates that our newly developed AI-based phenotypic screening system can effectively distinguish morphological changes induced by microbial broths and facilitates the identification of novel antifungal compounds.