Regularization of least squares problems in CHARMM parameter optimization by truncated singular value decompositions.

Area of Science:

• Computational Chemistry
• Molecular Mechanics
• Biophysics

Background:

• Molecular mechanics force field parameter optimization often involves ill-posed least squares problems.
• Accurate parameterization is crucial for studying complex systems like polycyclic aromatic hydrocarbon-DNA adducts.

Purpose of the Study:

• To apply truncated singular value decomposition and Tikhonov regularization to solve ill-posed least squares problems in force field optimization.
• To systematically determine parameters for accurate dihedral term optimization in molecular mechanics.

Main Methods:

• Utilized truncated singular value decomposition (SVD) and Tikhonov regularization.
• Applied methods to dihedral parameter optimization of genotoxic polycyclic aromatic hydrocarbon-DNA adducts.
• Employed discrete Picard condition and singular value spectrum analysis for parameter determination.

Main Results:

• Achieved systematic determination of truncation and regularization parameters.
• Produced unique, truncated, and regularized solutions for the ill-posed problems.
• Resulted in optimized force field dihedral terms that accurately parameterize the torsional energy landscape.

Conclusions:

• Truncated SVD and Tikhonov regularization offer an efficient approach to molecular mechanics force field parameter optimization.
• The unique solutions avoid the need for multiple iterations and trial-and-error.
• This method enhances the accuracy of force fields for studying chemical carcinogenesis.