Combined Machine Learning and Molecular Modelling Workflow for the Recognition of Potentially Novel Fungicides

Area of Science:

• Computational chemistry
• Drug discovery
• Fungal molecular biology

Background:

• Fungal infections pose a significant threat, necessitating the development of new antifungal agents.
• Drug repurposing offers a faster and more cost-effective approach to discovering novel therapeutics.
• Sterol biosynthesis pathways, involving targets like Cyp51 and Erg2, are crucial for fungal survival and represent promising targets for antifungal drugs.

Purpose of the Study:

• To identify novel fungicide targets within fungal sterol biosynthesis.
• To repurpose existing drugs as potential antifungal agents against these targets.
• To develop and validate predictive models for fungicide discovery.

Main Methods:

• Machine learning algorithms (FS-MLR, UVE-PLS, FS-LM-RF) were employed for molecular descriptor analysis.
• Molecular docking simulations (Autodock4, Gold) were performed to assess ligand-target interactions.
• Quantum mechanics/molecular mechanics (QM/MM) calculations were utilized for energy optimization and scoring of docked compounds.

Main Results:

• 112 prediction models were generated to identify potential fungicide hit compounds.
• Docking experiments identified several promising compounds targeting Cyp51 and Erg2.
• Seven lead compounds showed high scores for the Cyp51 target, and three for the Erg2 target based on QM/MM analysis.

Conclusions:

• The study successfully identified potential drug repurposing candidates for novel antifungal agents.
• Seven and three lead compounds were selected for Cyp51 and Erg2 targets, respectively.
• These identified compounds warrant further in vitro investigation for their antifungal efficacy.