Structure-guided compound prioritization strategy for virtual screening identifies putative binders for the nuclear...

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• Structure-based virtual screening often produces false positives due to pose variability and scoring biases.
• Accurate prioritization of potential drug candidates is crucial for efficient drug discovery.

Purpose of the Study:

• To develop an improved compound prioritization strategy for structure-based virtual screening.
• To enhance the identification of binders for the nuclear receptor LRH-1 (NR5A2).

Main Methods:

• Utilized sampled docked poses from physics-based and generative model docking approaches.
• Trained a multi-layer perceptron (MLP) model using contrasting docking results against multiple protein target models.
• Applied the MLP model to predict binders at the orthosteric ligand-binding pocket of LRH-1.

Main Results:

• The MLP model successfully identified known binders, including chemically dissimilar compounds and those with single scaffold modifications.
• A prospective virtual screening campaign using this strategy led to the discovery of four putative LRH-1 binders.
• Combining scoring and prediction metrics enriched hit compounds across various library sizes.

Conclusions:

• The developed strategy effectively leverages structural and experimental data to improve virtual screening for challenging targets like LRH-1.
• The MLP model shows potential as a tool for lead optimization in drug discovery.
• This approach offers a robust method to overcome limitations of single-pose scoring and ranking in virtual screening.