Polarizable Atomic Multipole X-Ray Refinement: Particle Mesh Ewald Electrostatics for Macromolecular Crystals

Area of Science:

• Structural Biology
• Computational Chemistry
• Biophysics

Background:

• Macromolecular model refinement in X-ray crystallography traditionally has limited support for rigorous molecular physics due to computational constraints.
• Electrostatic interactions are often neglected in favor of simpler nonbonded potentials, potentially limiting the accuracy of resulting structural models.
• Advancements in computational power and algorithms are needed to integrate sophisticated physical models into crystallography workflows.

Purpose of the Study:

• To present advanced algorithms enabling X-ray crystallography refinement using state-of-the-art classical molecular physics on desktop workstations.
• To implement and evaluate a symmetry-accelerated Particle Mesh Ewald (PME) summation for the polarizable Atomic Multipole Optimized Energetics for Biomolecular Applications (AMOEBA) force field.
• To demonstrate the benefits of incorporating AMOEBA electrostatics with PME for refining challenging macromolecular X-ray crystallography data.

Main Methods:

• Development of theory for PME summation that consistently handles all 230 space groups and unit cell replication.
• Implementation of a symmetry-accelerated PME engine (Force Field X - FFX) for the AMOEBA force field, optimized for multi-core workstations with GPU acceleration.
• Rerefinement of 10 crystallographic datasets (1.7–4.5 Å resolution) using AMOEBA with PME, comparing results to models refined without electrostatics.

Main Results:

• The FFX engine achieved over a 24-fold acceleration in energy evaluations compared to a single CPU core on a P1 unit cell.
• AMOEBA with PME refinement improved agreement with crystallographic data (lowered Rfree by 0.5%), enhanced geometric features (e.g., backbone conformations), and reduced average potential energy per residue (>10 kcal/mol).
• MolProbity validation indicated improved structural geometry, consistent with data up to 2.2 Å resolution, a 0.9 Å improvement over the mean quality.

Conclusions:

• Polarizable AMOEBA-assisted X-ray refinement offers significant advantages over methods neglecting electrostatics.
• The developed algorithms are efficient enough for routine application in macromolecular X-ray crystallography.
• This approach enhances the accuracy and quality of macromolecular models derived from X-ray diffraction data.