Development of an ab initio protein structure prediction system ABLE

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Protein structure prediction is crucial for understanding biological function.
• Conventional fragment assembly methods struggle with accurate native-like structure selection.
• Energy minimization alone is insufficient for identifying correct protein folds.

Purpose of the Study:

• To introduce ABLE, an ab initio protein structure prediction system.
• To improve native-like model selection in fragment assembly.
• To enhance the generation of diverse and accurate local protein structures.

Main Methods:

• ABLE utilizes a fragment assembly approach with a novel structural clustering method.
• Unit-vector root mean square distance (URMS) is employed for robust structure similarity measurement.
• Local structures are assigned using probability distribution maps of mainchain dihedral angles (phi, psi).
• Re-clustering and energy minimization with structural restraints are performed when initial clusters are inadequate.

Main Results:

• The ABLE system achieved near-native folds for more than half of the 25 small proteins tested.
• The structural clustering method demonstrated robust and effective selection of native-like models.
• The probability-based local structure assignment enabled generation of structures not present in existing databases.
• The developed structural clustering method shows potential applicability to other protein structure prediction systems.

Conclusions:

• ABLE offers an effective ab initio protein structure prediction strategy.
• Structural clustering significantly enhances the accuracy of fragment assembly methods.
• The system's ability to generate novel local structures improves prediction capabilities.
• The demonstrated methods hold promise for advancing the field of protein structure prediction.