AutoPH4: An Automated Method for Generating Pharmacophore Models from Protein Binding Pockets.

Area of Science:

• Computational chemistry
• Drug discovery
• Structural biology

Background:

• Pharmacophore models are crucial for computational drug discovery, aiding in virtual screening by defining ligand-target interactions.
• Traditional pharmacophore model generation is manual, limiting the integration of protein flexibility, often explored through molecular dynamics (MD) simulations.
• Automated methods are sought to process extensive protein conformational data and enhance virtual screening workflows.

Purpose of the Study:

• To introduce AutoPH4, an automated method for generating pharmacophore models directly from protein structures.
• To evaluate the performance of AutoPH4 within a virtual screening workflow.
• To demonstrate the potential of AutoPH4 in accurately ranking compounds and incorporating protein flexibility.

Main Methods:

• Development of AutoPH4, an automated algorithm for pharmacophore model generation from protein structures.
• Integration of AutoPH4 into a virtual screening workflow.
• Benchmarking the performance of the AutoPH4-based workflow against existing pharmacophore-based methods using public datasets.

Main Results:

• The virtual screening workflow incorporating AutoPH4 achieved superior compound ranking accuracy compared to all previously reported pharmacophore-based virtual screening workflows on a public benchmark.
• The high performance indicates that AutoPH4 generates high-quality pharmacophore models.
• AutoPH4 successfully facilitates the processing of multiple protein conformations, including those from MD simulations.

Conclusions:

• AutoPH4 represents a significant advancement in automated pharmacophore generation.
• The method enhances virtual screening accuracy, making it a valuable tool for drug discovery.
• AutoPH4 is well-suited for future virtual screening applications, particularly those involving dynamic protein structures from MD simulations.