Employing Machine Learning Techniques to Detect Protein Function: A Survey, Experimental, and Empirical Evaluations.

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• Protein function prediction is crucial for understanding biological systems.
• Diverse machine learning (ML) algorithms exist, but a unified classification is lacking.
• Evaluating and comparing these ML techniques is essential for advancing the field.

Purpose of the Study:

• To comprehensively review machine learning methods for protein function identification.
• To propose an innovative hierarchical classification system for ML algorithms.
• To empirically and experimentally evaluate various ML techniques.

Main Methods:

• Development of a tri-level hierarchical classification system for ML algorithms.
• Empirical evaluation of techniques based on four criteria.
• Experimental assessment of individual techniques, sub-categories, and broad categories.
• Inclusion of single-task, multi-task, and multi-label function detection methods.

Main Results:

• The hierarchical classification provides a structured framework for understanding algorithm interrelationships.
• Empirical and experimental evaluations differentiate technique efficacy and performance.
• The study offers a well-rounded understanding of ML strategies in protein function identification.

Conclusions:

• The proposed classification and evaluation framework enhance the understanding of ML in protein function prediction.
• Future research directions for ML in this domain are highlighted.
• This work aids researchers in selecting appropriate ML tools for protein function determination.