Artificial Intelligence‐Guided Design and Computational Evaluation of Candidate CYP51B Inhibitors as Potential Antifungal Agents

Résumé fourni par la source

ABSTRACT CYP51B (sterol 14α‐demethylase) is essential for ergosterol biosynthesis in Aspergillus fumigatus and is the target of the clinical azole antifungals. A curated library of 174 antifungal agents was expanded with the Mol2Mol module of REINVENT4 to 111,655 unique structures, of which 60,120 survived diversity filtering and 6910 passed ADMET‐AI screening. Docking into CYP51B (PDB 4UYM) with PLANTS returned 1811 compounds scoring better than the co‐crystallized ligand voriconazole (VOR); DeepSA and K Deep filtering gave 26, from which five candidates (AFR1–AFR5) were selected. Fingerprint analysis shows that these span a wide novelty range, from AFR3 (maximum Tanimoto coefficient 0.32) to close analogues of efinaconazole (AFR4, 0.88) and fluconazole (AFR5, 0.79). Each complex, with the CYP51B–VOR reference, was simulated in duplicate for 100 ns from independent velocity seeds. All complexes remained stable, with replicate‐averaged backbone RMSD of 0.20–0.24 nm. Replicate‐averaged MM‐GBSA effective energies were −99.04 (AFR2), −95.06 (AFR5), −90.83 (VOR), −90.41 (AFR4), −85.36 (AFR1) and −84.31 kcal/mol (AFR3); because the replicate spread of a single system is comparable to the differences between ligands, these are relative rankings rather than absolute affinities. AFR2 is proposed for experimental follow‐up. All findings are computational predictions requiring experimental validation.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.