Optimization of Cavity-Based Negative Images to Boost Docking Enrichment in Virtual Screening.

Area of Science:

• Computational chemistry
• Drug discovery and development
• Bioinformatics

Background:

• Molecular docking is crucial for drug discovery but often struggles to differentiate active from inactive compounds.
• Existing negative image-based rescoring (R-NiB) methods improve docking yield but require expert manual editing for optimal performance.

Purpose of the Study:

• To present a novel, automated methodology, brute force negative image-based optimization (BR-NiB), for optimizing protein ligand binding cavity models.
• To enhance the accuracy and efficiency of virtual screening campaigns in drug discovery.

Main Methods:

• Developed a greedy search-driven methodology (BR-NiB) for iterative editing and benchmarking of cavity models.
• Trained, tested, and validated BR-NiB rigorously across multiple drug targets and docking software.
• Compared BR-NiB to existing R-NiB methods, highlighting its automated optimization capabilities.

Main Results:

• BR-NiB significantly improves the efficacy of molecular docking by optimizing shape-focused pharmacophore models.
• The optimized models effectively filter docked active compounds from inactive or decoy compounds.
• BR-NiB ensures excellent docking performance, validated across diverse drug targets and docking tools.

Conclusions:

• BR-NiB represents a new approach to shape-focused pharmacophore modeling, automating the optimization of cavity information.
• This method boosts the success rates of docking-based virtual screening campaigns.
• The BR-NiB code is freely available on GitHub to facilitate its adoption in drug discovery research.