TP-ML: A Machine-Learning-Based Tool to Identify Threonine Proteases Using Sequence-Derived Optimal Features.

Area of Science:

• Biochemistry and Bioinformatics
• Enzymology
• Computational Biology

Background:

• Threonine proteases (TPs) are crucial enzymes involved in biological processes and diseases like Alzheimer's and cancer.
• Their intracellular protein degradation capabilities offer therapeutic and industrial potential.
• Current experimental methods for TP identification are laborious and costly.

Purpose of the Study:

• To develop an efficient computational tool for predicting threonine proteases (TPs).
• To accelerate the identification and characterization of novel TPs for research and applications.
• To overcome the limitations of traditional experimental TP discovery methods.

Main Methods:

• A support vector machine-based prediction tool, TP-ML, was developed.
• A benchmark dataset was created, and physicochemical and compositional features were extracted from amino acid sequences.
• Feature selection approaches were compared to identify optimal feature sets.
• Five machine-learning classifiers were trained and evaluated.

Main Results:

• TP-ML demonstrated robust performance, selected as the best model.
• The tool achieved high accuracy in both cross-validation (0.934) and independent evaluation (0.888).
• The study identified optimal feature sets for TP prediction.

Conclusions:

• TP-ML is a powerful computational tool for identifying threonine proteases.
• This tool can significantly aid in the experimental characterization of TPs.
• TP-ML facilitates exploration of industrial applications for threonine proteases.