Enhancing protein function prediction with taxonomic constraints--The Argot2.5 web server.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein function prediction is crucial for understanding biological systems.
• Existing tools like Argot2 require updates for improved performance and usability.
• Gene Ontology (GO) terms provide a standardized framework for describing protein functions.

Purpose of the Study:

• To introduce Argot2.5, an enhanced web server for predicting protein function.
• To improve the usability and performance of the Argot2 tool.
• To leverage semantic similarity of Gene Ontology terms and taxonomic constraints for more accurate predictions.

Main Methods:

• Exploited grouping of Gene Ontology terms via semantic similarity for function inference.
• Implemented input data selection from clustered UniProt databank searches and remodeled Pfam hit weights.
• Applied taxonomic constraints using the FunTaxIS tool to filter species-specific annotations.

Main Results:

• Argot2.5 demonstrated superior performance compared to Argot2, PANNZER, and a BLAST baseline on two independent benchmarks.
• Key interventions, including data selection and taxonomic filtering, significantly contributed to performance gains.
• The updated algorithmic strategy and feature enhancements improved overall usability and predictive accuracy.

Conclusions:

• Argot2.5 represents a significant advancement in protein function prediction tools.
• The integration of semantic similarity and taxonomic filtering enhances prediction accuracy and reliability.
• The web server is freely available for academic use, facilitating research in bioinformatics and computational biology.