From AI modelling to validation: A 10-day pipeline for potential lead compounds validation

Fusarium graminearum, one of the dominant causal agents of Fusarium head blight (FHB), not only causes significant yield losses but also contaminates grains by secreting mycotoxins such as deoxynivalenol (DON). Haem, a cofactor for multiple enzymes, requires coproporphyrinogen III oxidase (Cpo) as a rate-limiting enzyme for its biosynthesis. Although the knockout of FgCpo was lethal in F. graminearum, its potential role as a fungicide target remains largely unexplored, and there are no studies on the screening of inhibitors against this protein. Therefore, in this study, we established an efficient 10-day integrated pipeline for rapid screening and validation of FgCpo-targeted antifungal inhibitors. This pipeline comprised four streamlined phases: (i) Protein modelling (1 h): A high-precision FgCpo 3D model was generated using AlphaFold3; (ii) Virtual screening (4 days): A dual-docking strategy screened 1.6 million compounds, identifying 8 high-affinity candidates; (iii) Molecular dynamics simulations (3 days): Revealed competitive binding of Chemdiv_2925 at the Gly357/Thr358 sites; (iv) Bioassays (3-5 days): The EC₅₀ of Chemdiv_2925 against 8 phytopathogenic fungi ranged from 23.55 μg/mL to 198.83 μg/mL, with a MIC value of 41 μg/mL against spore germination of F. graminearum, along with a 69.6 % inhibition of deoxynivalenol (DON). Relative gene expression analysis revealed that the expression of haem synthesis pathway genes was significantly upregulated under Chemdiv_2925 treatment conditions. The entire procedure-from target modelling to compound validation-was completed within 10 days, establishing an innovative approach for facilitating the rapid development of highly efficient and low-toxicity lead compounds for pesticide discovery and development.