Improving the accuracy of template-based predictions by mixing and matching between initial models

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Comparative modeling predicts protein structures using known structures.
• Refining initial models to experimental resolution remains a challenge.
• Current methods struggle with accurate model refinement.

Purpose of the Study:

• To investigate a graph-theoretic clique finding approach for refining comparative models.
• To address the bottleneck in improving the resolution of initial protein structure predictions.
• To enhance the accuracy of protein structure prediction through advanced computational methods.

Main Methods:

• Utilized a graph-theoretic clique finding algorithm.
• Developed a method for mixing and matching regions from multiple initial comparative models.
• Applied the approach to refine protein conformations in three dimensions.

Main Results:

• The novel method integrates information from multiple templates/alignments.
• Optimized conformation ensembles with improved secondary structures were generated.
• Refined models demonstrated higher accuracy compared to initial models in CASP7.
• Near-native conformations were effectively accumulated and identified.

Conclusions:

• The approach offers an automated method for improving comparative modeling predictions.
• It enhances conformational model quality when multiple template/alignment combinations are available.
• This technique provides higher quality conformational models than initial predictions without manual intervention.