Rapid model quality assessment for protein structure predictions using the comparison of multiple models without...

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Structure Prediction

Background:

• Accurate prediction of 3D protein model quality is crucial for tertiary structure prediction.
• Current leading methods (MQAPs) are accurate but computationally intensive due to structural alignments.
• Existing methods include clustering- and consensus-based approaches.

Purpose of the Study:

• Introduce ModFOLDclustQ, a novel Model Quality Assessment Program (MQAP).
• Develop ModFOLDclust2 by combining ModFOLDclustQ with the existing ModFOLDclust method.
• Enhance prediction accuracy and speed for 3D protein models.

Main Methods:

• ModFOLDclustQ utilizes the Q measure for model comparisons, avoiding CPU-intensive structural alignments.
• Benchmarking ModFOLDclustQ against established methods for accuracy and speed.
• Integrating ModFOLDclustQ scores with ModFOLDclust scores to create ModFOLDclust2.

Main Results:

• ModFOLDclustQ is competitive with leading MQAPs in global model quality prediction.
• ModFOLDclustQ is significantly faster than previous methods: up to 150x for small proteins and 5x for large proteins.
• ModFOLDclust2 shows improved accuracy over previous MQAPs with negligible computational overhead.

Conclusions:

• ModFOLDclustQ offers a faster and accurate alternative for 3D protein model quality assessment.
• ModFOLDclust2 provides enhanced prediction accuracy by combining novel and existing scoring methods.
• These methods are available for download, facilitating their use in protein structure prediction research.