Reduced representation model of protein structure prediction: statistical potential and genetic algorithms

Area of Science:

• Computational Biology
• Structural Biology
• Bioinformatics

Background:

• Protein structure prediction is crucial for understanding protein function.
• Accurate prediction requires efficient computational methods.
• Reduced representation models simplify complex protein structures.

Purpose of the Study:

• To develop and validate a novel computational method for protein structure prediction.
• To assess the accuracy of the method using small proteins with known crystal structures.
• To explore the application of genetic algorithms in protein folding.

Main Methods:

• Utilized a reduced representation model focusing on backbone dihedral angles (phi and psi).
• Employed a statistical potential function incorporating local and nonlocal interactions.
• Developed a genetic algorithm to optimize a population of conformations simultaneously.
• Initiated simulations from random conformations using only primary sequence and radius of gyration.

Main Results:

• Successfully predicted folded structures for melittin (26 residues) with high computational convergence.
• Achieved an average root mean square error of 1.66 Å compared to crystal structures.
• Obtained similar accurate results for avian pancreatic polypeptide inhibitor (36 residues).
• Demonstrated accurate folding and correct disulfide bond formation for apamin (18 residues).

Conclusions:

• The developed genetic algorithm-based method is effective for predicting native-like protein structures.
• The approach shows high accuracy and computational efficiency for small proteins.
• This method holds promise for advancing protein structure prediction in bioinformatics.