PL-PatchSurfer3: improved structure-based virtual screening for structure variation using 3D Zernike descriptors

Area of Science:

• Computational chemistry
• Drug discovery
• Structural bioinformatics

Background:

• Structure-based virtual screening (SBVS) is crucial for in silico drug discovery.
• SBVS performance is highly dependent on the accuracy of the receptor's 3D structure.
• Existing methods often fail when screening apo structures against holo targets due to conformational changes.

Purpose of the Study:

• To introduce PL-PatchSurfer3, an enhanced SBVS method.
• To improve robustness and accuracy in virtual screening.
• To address the limitations of conventional docking methods concerning protein conformational flexibility.

Main Methods:

• Utilized a surface patch-based representation of binding sites and ligands.
• Employed 3D Zernike descriptors to capture shape and physicochemical properties of molecular surfaces.
• Incorporated refined hydrogen bond complementarity and visibility (curvature) for enhanced patch description.

Main Results:

• PL-PatchSurfer3 demonstrated superior performance compared to its predecessor, PL-PatchSurfer.
• The method showed robustness across diverse receptor structures, including apo, holo, modeled, and AlphaFold-predicted forms.
• PL-PatchSurfer3 outperformed or matched conventional and recent deep learning-based SBVS methods.

Conclusions:

• PL-PatchSurfer3 offers a more stable and accurate SBVS approach.
• The surface patch methodology effectively handles receptor conformational variations.
• This advancement holds significant potential for improving the efficiency of in silico drug discovery pipelines.