Ab initio folding of proteins using restraints derived from evolutionary information.

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Folding

Background:

• The ab initio protein structure prediction challenge is crucial for understanding protein function.
• Accurate prediction of protein three-dimensional structures from amino acid sequences remains a significant computational hurdle.

Purpose of the Study:

• To evaluate a novel ab initio protein folding method in the CASP3 (Critical Assessment of protein Structure Prediction) competition.
• To assess the method's accuracy in predicting global folds and topology for various protein targets.

Main Methods:

• Utilized a Monte Carlo search algorithm.
• Incorporated secondary and tertiary restraints derived from multiple sequence alignments.
• Applied the method to eleven protein targets in the CASP3 experiment.

Main Results:

• Achieved "correct" global folds for targets 65 and 74, and "almost correct" for targets 64, 75, and 77.
• Successfully predicted low-resolution structures with good global topology for small proteins (≤110 residues).
• Demonstrated potential for predicting rare or novel folds, particularly for targets 75 and 77.

Conclusions:

• The ab initio folding method shows progress, especially for small proteins, offering comparable or better results than threading methods.
• Limitations include difficulties with beta proteins, proteins exceeding 110 residues, and poor multiple sequence alignments.
• The approach advances the field of ab initio protein folding, offering an alternative to structure databases.