Prediction of protein structure from ideal forms.

Area of Science:

• Computational Biology
• Structural Bioinformatics
• Protein Folding

Background:

• The sequence of a protein is known to determine its 3D structure.
• Predicting protein structure computationally from sequence alone has been a long-standing challenge, especially for proteins larger than 100 residues.
• Existing computational methods often rely on sequence homology or structural templates.

Purpose of the Study:

• To develop a novel computational method for predicting protein three-dimensional structures.
• To accurately predict complex protein folds for proteins up to 200 residues without using sequence homology or specific structural data.
• To provide an alternative to template-based or fragment-based prediction methods.

Main Methods:

• A computational approach generating thousands of models based on idealized structural representations.
• Scoring and refinement of generated models.
• Utilizing a database of ideal models derived from general packing rules, akin to an ab initio approach.

Main Results:

• Successfully predicted complex protein folds for proteins up to nearly 200 residues.
• The correct fold was ranked highest for proteins over 150 residues in test sets.
• Achieved high scores for correct folds on a set of five proteins.
• Identified close topological variants for some larger proteins.

Conclusions:

• The developed method offers a significant advancement in predicting protein structure from sequence.
• It provides a viable computational strategy for larger proteins without relying on homologous structures.
• The approach represents a step towards true ab initio protein structure prediction.