Evaluation of protein fold comparison servers

Area of Science:

• Structural bioinformatics
• Computational biology
• Protein science

Background:

• Protein structure determination is advancing rapidly, necessitating robust methods for classifying newly discovered protein folds.
• Understanding protein fold classification aids in inferring protein function and evolutionary relationships.
• Existing automatic fold comparison programs lack comprehensive comparative analysis.

Purpose of the Study:

• To critically evaluate the performance and functionality of 11 publicly available web servers for automatic protein fold comparison.
• To identify the strengths and weaknesses of different fold comparison methodologies.
• To provide recommendations for researchers utilizing these tools.

Main Methods:

• Assessed 11 web-based automatic fold comparison servers on functionality (user interface, results presentation) and performance (recognition of structural similarities).
• Conducted extensive performance tests using a diverse dataset including standard folds, multidomain proteins, Calpha-only models, novel folds, and NMR structures.
• Utilized the CATH structural classification system as a benchmark for evaluating server accuracy.

Main Results:

• Significant variations in performance and functionality were observed across the evaluated servers.
• CE, DALI, MATRAS, and VAST demonstrated superior performance in recognizing established structural similarities.
• No single server achieved a 100% success rate in identifying all structural relationships within the test dataset.

Conclusions:

• The evaluation revealed distinct strengths and limitations for each fold comparison server.
• Researchers should consider using multiple servers to increase the confidence in fold classification and novelty assessment.
• Complementary use of different servers is recommended before concluding the novelty of a protein fold.