A multilayer evaluation approach for protein structure prediction and model quality assessment.

Area of Science:

• Computational biology
• Structural bioinformatics
• Molecular modeling

Background:

• Protein tertiary structures are crucial for understanding molecular functions.
• Computational prediction is vital for protein structure determination.
• Current quality assessment (QA) tools often fail to identify the best models from predicted pools due to incomplete understanding of protein folding.

Purpose of the Study:

• To develop novel QA methods to enhance protein structure prediction.
• To reveal new structural facets complementary to existing QA methods.
• To integrate diverse QA approaches into a unified system for improved model selection.

Main Methods:

• Development of two QA methods for target/template alignments.
• Implementation of a molecular dynamics (MD)-based QA method.
• Creation of three consensus QA methods and analysis of inter-method relationships.
• Integration into a multilayer evaluation approach within the MUFOLD system.

Main Results:

• The new QA methods provide complementary insights into protein structure models.
• Analysis revealed underlying relationships among different QA methods.
• The integrated multilayer evaluation approach guides model generation and selection.
• The MUFOLD system demonstrated improved QA discerning power and prediction accuracy in CASP8 and CASP9.

Conclusions:

• Novel QA methods and a multilayer evaluation approach significantly improve protein structure prediction.
• The MUFOLD system offers enhanced accuracy in selecting high-quality protein models.
• This work advances computational approaches for determining protein tertiary structures.