A structural pattern-based method for protein fold recognition.

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein structure prediction

Background:

• Assessing sequence-structure relationships is crucial for understanding protein function.
• Existing methods for evaluating sequence-structure alignment can be computationally intensive.
• The Protein Data Bank (PDB) provides a valuable resource for structural pattern analysis.

Purpose of the Study:

• To develop and evaluate a novel method, SPREK, for assessing sequence-structure register.
• To compare SPREK's performance against existing computational approaches using benchmark datasets.
• To identify limitations and strengths of SPREK across various protein sizes and complexities.

Main Methods:

• SPREK method development based on matching structural patterns against a PDB-derived library.
• Score normalization using sequence shuffling and permutation to establish random background distributions.
• Evaluation using random-walk and sophisticated structure prediction generated 'random' structures.
• Performance assessment on decoy models and CASP-5 exercise models.

Main Results:

• SPREK performance on decoy models was equivalent or superior to more complex methods.
• SPREK showed limitations with very small proteins due to insufficient patterns.
• For intermediate CASP targets, SPREK ranked half of the models within the top 10%.
• Suboptimal rankings occurred when SPREK selected a well-packed model with an incorrect fold, or for very large/small proteins.

Conclusions:

• SPREK offers an effective approach for evaluating sequence-structure register, particularly for medium-sized proteins.
• The method demonstrates competitive performance against established techniques.
• Further refinement may be needed to address challenges with extremely small or large protein structures.