Local feature frequency profile: a method to measure structural similarity in proteins

Area of Science:

• Structural Bioinformatics
• Computational Biology
• Protein Structure Analysis

Background:

• Protein structure classification and understanding the protein structure universe rely on accurate structural similarity measures.
• Existing methods for assessing protein structural similarity can be computationally intensive.

Purpose of the Study:

• To develop a rapid and efficient method for measuring structural similarity between protein structures.
• To create a new representation for protein structures that facilitates large-scale analysis and visualization of the protein structure universe.

Main Methods:

• Extraction of representative local feature (LF) patterns from C(alpha) distance matrices of protein structures using medoid analysis.
• Encoding protein structures by labeling submatrices based on the nearest representative LF patterns.
• Representing each protein structure as a local feature frequency (LFF) profile, which is a frequency distribution of these labels.

Main Results:

• The LF frequency (LFF) profile method enables quick calculation of structural similarity scores for numerous protein structures.
• This approach facilitates the construction and updating of a comprehensive map of the protein structure universe.
• The LFF profile method effectively maps complex protein structures into a common Euclidean space without requiring secondary structure information or structural alignment.

Conclusions:

• The LFF profile method offers a computationally efficient and objective approach to protein structure comparison and classification.
• This method provides a valuable tool for exploring the global landscape of protein structures.
• The technique simplifies the representation of complex protein structures for large-scale bioinformatics analyses.