Protein function prediction with gene ontology: from traditional to deep learning models

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Protein function prediction is vital for genome annotation.
• High-throughput sequencing drives rapid development in prediction methods.
• Gene Ontology (GO) is a key resource for describing protein functions.

Purpose of the Study:

• To review computational Gene Ontology (GO) annotation methods for proteins.
• To compare the performance of selected GO prediction tools.
• To discuss challenges and future directions in protein function prediction.

Main Methods:

• Comprehensive review of existing computational GO annotation methods.
• Categorization of methods into conventional and deep learning approaches.
• Performance evaluation of selected predictors using a benchmark dataset.

Main Results:

• Deep learning methods show significant potential for GO term assignment.
• A comparative analysis of selected tools was performed on a challenge dataset.
• Identified key challenges and promising future research avenues.

Conclusions:

• Computational methods, especially deep learning, are advancing protein function prediction.
• Further research is needed to address current limitations in GO annotation.
• Future directions focus on improving accuracy and efficiency in predicting protein functions.