Nonbonded terms extrapolated from nonlocal knowledge-based energy functions improve error detection in near-native...

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Accurate protein structure assessment is crucial for predicting protein structures, refining models, and inferring function.
• Knowledge-based potentials are widely used for evaluating protein models.

Purpose of the Study:

• To assess and compare the effectiveness of different full atom knowledge-based potentials in detecting small, localized errors in protein models.
• To evaluate the impact of incorporating close nonbonded pairwise atom terms on classifying residue modeling accuracy.

Main Methods:

• Compared various full atom knowledge-based potentials for error detection in comparative protein models.
• Investigated the effect of close nonbonded pairwise atom terms on accuracy assessment.
• Extrapolated close nonbonded terms from pseudo-energy functions of nonlocal potentials due to limitations in direct experimental derivation.

Main Results:

• The proposed methodology, incorporating extrapolated close nonbonded terms, significantly improved the detection of errors in protein models.
• Directly derived nonbonded terms from experimental data showed poor performance.
• Some widely used external knowledge-based energy functions also performed poorly on this specific error detection task.

Conclusions:

• A proper description of close nonbonded terms is essential for accurately representing native protein conformations.
• The developed methodology enhances the detection of localized structural errors in both modeled and experimental protein structures.
• This approach offers a valuable tool for improving the quality assessment of protein structures.