De Novo Discovery of Ecofriendly Antifungal Agents Using Generative AI and Hierarchical Virtual Screening

The discovery of next-generation agrochemicals is limited by analogue-based molecular design and the high cost of experimental screening. Here, we utilized a computational workflow combining generative AI, pharmacophore screening, hierarchical molecular docking, and protein-ligand binding compatibility assessment to discover CYP51-targeting antifungal agents. Applied to CYP51 from Rhizoctonia solani, this workflow prioritized thiazole-triazole hybrids, among which compound F50 showed strong antifungal activity. F50 exhibited potent in vitro and in vivo activity against R. solani and field efficacy comparable to or slightly better than tebuconazole under the tested conditions. Preliminary biosafety evaluation indicated lower acute toxicity of F50 toward zebrafish, earthworms, and honeybees than that of tebuconazole. Molecular docking and dynamics simulations suggested that the thiazole-containing scaffold and p-chloro substituent contributed to stable F50 binding within the CYP51 pocket. These results demonstrate the utility of combining generative molecular design with hierarchical virtual screening to identify promising antifungal lead compounds for sustainable crop protection.