Unleashing the power of meta-threading for evolution/structure-based function inference of proteins

Area of Science:

• Computational biology
• Structural bioinformatics
• Protein science

Background:

• Protein threading is crucial for predicting protein structure and function.
• Current methods often use similar criteria for structure and function template identification, leading to high false positive rates in function inference.
• Selecting functional templates requires distinct constraints beyond mere structural similarity.

Purpose of the Study:

• To extend the eThread meta-threading approach for optimal functional template selection.
• To cover a broad spectrum of protein molecular functions, including ligand, metal, inorganic cluster, protein, and nucleic acid binding.
• To improve the accuracy and reduce false positives in protein function annotation.

Main Methods:

• Optimized template selection within the eThread meta-threading framework.
• Large-scale benchmarking to evaluate template recognition rates and false positive rates.
• Comparative analysis of meta-threading against single-threading approaches.
• Machine learning integration to enhance functional template selection.

Main Results:

• High recognition rates (70-80%) for identifying templates binding molecular partners in similar locations with low false positives.
• Demonstrated high sensitivity (58%) in recognizing similar protein-binding interfaces at an 18% false positive rate.
• Meta-threading with machine learning significantly outperforms single-threading for functional template selection.
• Effective detection of diverse protein molecular functions, even in low-sequence identity scenarios.

Conclusions:

• The enhanced eThread approach provides accurate and reliable functional template selection.
• Meta-threading is superior to single-threading for inferring protein molecular function.
• eThread offers valuable insights into binding molecule properties and structural features.
• The enhanced eThread is available as a webserver and stand-alone software.