Polarizable protein packing.

Area of Science:

• Computational chemistry
• Protein structure prediction
• Molecular modeling

Background:

• Protein combinatorial optimization requires accurate energy calculations.
• Polarizable force fields like AMOEBA are computationally expensive.
• Approximating complex energy functions is crucial for efficiency.

Purpose of the Study:

• To incorporate protein polarization effects into combinatorial optimization.
• To develop a computationally tractable approximation of the AMOEBA force field.
• To enable efficient identification of low-energy protein conformations.

Main Methods:

• Decomposition of the AMOEBA force field into low-order terms (up to third-order).
• Representation of the protein packing problem as a hypergraph.
• Solving the optimization problem using the FASTER algorithm.
• Improving approximate energy models using ridge regression.

Main Results:

• Third-order approximation of AMOEBA provides a tractable energy model.
• The FASTER algorithm efficiently finds optimal rotamers.
• Ridge regression achieves high accuracy (rmsd < 1 kJ mol(-1)) for trained models.
• The method successfully identifies new, low-energy protein solutions.

Conclusions:

• Low-order approximations of polarizable force fields are effective for protein optimization.
• The developed framework enables efficient and accurate identification of protein structures.
• This general approach can be applied to other many-body combinatorial optimization problems.