Support vector machine prediction of enzyme function with conjoint triad feature and hierarchical context

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Enzymes, the largest protein class, are classified by the Enzyme Commission (EC) hierarchy.
• Accurate enzyme function categorization is vital for understanding molecular mechanisms.

Purpose of the Study:

• To develop an efficient and accurate machine learning framework for predicting enzyme function within the EC hierarchy.
• To improve protein sequence representation and incorporate structural information for enhanced prediction.

Main Methods:

• Proposed conjoint triad features (CTF) for protein sequence encoding, capturing amino acid composition and neighbor relationships.
• Developed a support vector machine-based method (SVMHL) that leverages the hierarchical structure of EC classifications.
• Implemented a structure-based prediction approach with low computational complexity.

Main Results:

• SVMHL with CTF significantly outperformed SVMHL with amino acid composition (AAC) in predictive accuracy and Matthew's correlation coefficient (MCC).
• Achieved high accuracy (81%-98%) and MCC (0.82-0.98) for predicting the first three EC digits on a low-homologous dataset.
• Demonstrated superior performance compared to methods ignoring hierarchical relationships or using prior inter-class knowledge.

Conclusions:

• The proposed SVMHL with CTF model offers reduced computational complexity and superior performance in enzyme function prediction.
• This novel method serves as a valuable tool for the enzyme function prediction community.