A Comparison between Enrichment Optimization Algorithm (EOA)-Based and Docking-Based Virtual Screening.

Area of Science:

• Computational Chemistry
• Drug Discovery
• Bioinformatics

Background:

• Virtual screening (VS) is crucial for drug and material design, utilizing structure-based (e.g., molecular docking) or ligand-based methods.
• Existing ligand-based methods often require project-specific models and performance metrics not optimized for VS.
• Current metrics for model evaluation are limited in assessing performance specifically within the context of VS.

Purpose of the Study:

• To present an improved Enrichment Optimization Algorithm (EOA) for virtual screening (VS).
• To enhance EOA's handling of active and inactive compounds for more robust predictive modeling.
• To compare the performance of the improved EOA against established molecular docking tools in VS experiments.

Main Methods:

• Developed an improved version of the Enrichment Optimization Algorithm (EOA).
• EOA generates Quantitative Structure-Activity Relationship (QSAR) models using multiple linear regression (MLR).
• Compared EOA against Glide-SP, GOLD, and AutoDock Vina using five molecular targets in small-scale VS experiments.

Main Results:

• The improved EOA consistently outperformed Glide-SP, GOLD, and AutoDock Vina in VS.
• EOA achieved superior performance based on Area Under the ROC Curve (AUC) and Enrichment Factor at 1% (EF1%) metrics.
• Performance advantages were observed regardless of whether docking metrics used a consensus approach or single crystal structures.

Conclusions:

• The improved EOA offers superior performance for virtual screening compared to traditional docking tools.
• EOA effectively utilizes information from both active and inactive compounds for enhanced predictive accuracy.
• Combining EOA with molecular docking presents a promising strategy for developing target-specific scoring functions.