Protein structure mining using a structural alphabet

Area of Science:

• Bioinformatics
• Structural Biology
• Computational Biology

Background:

• Protein structure comparison is crucial for understanding protein function and evolution.
• Existing methods often rely heavily on secondary structure elements, limiting their ability to capture local structural features.
• There is a need for efficient and accurate methods for mining structural similarities in large protein databases.

Purpose of the Study:

• To introduce and evaluate PB-ALIGN, a novel protein structure mining method.
• To assess the performance of PB-ALIGN in identifying protein folds and superfamilies.
• To compare PB-ALIGN with existing structure comparison methods.

Main Methods:

• Encoding protein structures as 1D sequences of 16 protein blocks (PBs).
• Utilizing a PB substitution matrix and dynamic programming for sequence alignment.
• Aligning local structural features beyond regular secondary structures.

Main Results:

• PB-ALIGN achieved 85% efficiency in extracting true folds from the SCOP database.
• Successfully identified true superfamily members in 82% of cases.
• Outperformed 13 existing methods on general datasets and matched top methods on nontrivial datasets, with faster processing speeds.

Conclusions:

• PB-ALIGN is an efficient and accurate method for protein structure mining.
• The method effectively captures local structural features, including loops and secondary structure caps.
• PB-ALIGN offers a reliable approach for fold demarcation and is available as a web server (PBE).