Automatic Gene Function Prediction in the 2020's

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• The rapid generation of DNA and protein sequences outpaces experimental functional discovery.
• Accurate Automatic Function Prediction (AFP) is crucial for understanding biological data.
• Despite progress, AFP remains an unsolved challenge in bioinformatics.

Purpose of the Study:

• To identify key challenges hindering the advancement of AFP.
• To propose strategies for improving AFP applicability and performance.
• To highlight the potential of machine learning in future AFP development.

Main Methods:

• Discussing challenges in condition-specific functional annotation.
• Addressing the prediction of functions for non-model organisms.
• Incorporating novel data sources and mitigating Gene Ontology (GO) annotation biases.
• Optimizing the representation of proteins, genes, and functions for prediction algorithms.

Main Results:

• Identifying five critical challenges in current AFP research.
• Recommending adaptations in data representation and algorithmic approaches.
• Demonstrating the need for continued machine learning integration in AFP.

Conclusions:

• AFP is a dynamic research area with significant potential for advancement.
• Overcoming identified challenges will enhance the utility of AFP.
• Machine learning innovations are key to the next generation of AFP in computational biology.