Protein side chain conformation predictions with an MMGBSA energy function

Area of Science:

• Structural Biology
• Computational Biology
• Biophysics

Background:

• Protein side chain conformation prediction is vital for protein structure prediction and design.
• The combinatorial complexity of this task presents a significant challenge.
• Molecular mechanics combined with Generalized Born and Surface Area (GBSA) models offer potential solutions.

Purpose of the Study:

• To evaluate the performance of the MMGBSA energy function within the Proteus protein design program for side chain conformation prediction.
• To assess the impact of individual energy terms (GB and SA) and parameterizations.
• To compare the accuracy of Proteus against specialized side chain prediction software.

Main Methods:

• The study employed the MMGBSA energy function implemented in the Proteus program.
• Side chain conformations were predicted for 18 proteins, both individually and collectively.
• Energy terms were evaluated, parameterizations compared, and side chain minimization was utilized.

Main Results:

• Generalized Born (GB) and Surface Area (SA) terms improved single side chain predictions with optimal parameters.
• For collective side chain prediction, pairwise additive approximation errors negated the benefits of GB and SA terms.
• Side chain minimization significantly improved accuracy by mitigating the rigid rotamer approximation.

Conclusions:

• MMGBSA with appropriate parameterization can achieve accuracies comparable to specialized tools like SCWRL4, particularly for core residues.
• The pairwise additive approximation is a limitation for collective side chain prediction.
• Minimization is essential for accurate side chain conformation prediction, even without GB/SA terms.