Improving the quality of protein structure models by selecting from alignment alternatives

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Model Quality Assessment Programs (MQAPs) are crucial for distinguishing high-quality protein structure models.
• Developing effective MQAPs is an active area of research in protein structure prediction.
• This study proposes a novel method to leverage MQAPs for improving protein model quality.

Purpose of the Study:

• To introduce a new method for enhancing protein structure model quality using MQAPs.
• To optimize model selection by systematically varying alignments and scoring them with an MQAP.
• To utilize a Support Vector Machine (SVM)-based scheme for combining MQAP potentials.

Main Methods:

• Generating multiple sequence alignments and corresponding protein models for a target sequence using a template structure.
• Scoring the quality of generated models using an MQAP.
• Employing an SVM-based selection scheme to choose the most promising model based on MQAP scores.
• Validating the method's effectiveness by comparing selected structures to native structures using TM-score.

Main Results:

• The proposed method demonstrates a significant increase in protein model quality.
• Statistical analysis (Wilcoxon signed rank test) shows p-values below 10(-15), confirming the significance of the improvement.
• The average increase in TM-score was 0.016, with a maximum observed increase of 0.29.

Conclusions:

• Alignment is a critical bottleneck in template-based protein structure prediction.
• Combining systematic alignment variation with modern model scoring functions substantially enhances the quality of alignment-based models.
• The developed method offers a promising approach for improving protein structure prediction accuracy.