Classifying nitrilases as aliphatic and aromatic using machine learning technique.

Area of Science:

• Biochemistry and Bioinformatics
• Computational Biology
• Enzyme Classification

Background:

• Nitrilases are enzymes that catalyze the hydrolysis of nitriles.
• Classifying nitrilases as aliphatic or aromatic is crucial for understanding their function and applications.
• Existing classification methods may lack efficiency or require extensive experimental data.

Purpose of the Study:

• To develop and validate a machine learning approach for classifying nitrilases.
• To differentiate between aliphatic and aromatic nitrilases using sequence-based features.
• To provide a computational tool for predicting nitrilase classification.

Main Methods:

• Utilized the ProCos (Protein Composition Server) machine learning technique.
• Employed feature vectors including pseudo-amino acid composition (PAAC) and five-factor solution score (5FSS).
• Trained and evaluated the algorithm on a dataset of known nitrilases.

Main Results:

• Pseudo-amino acid composition (PAAC) achieved 95.00% accuracy, 100.00% sensitivity, 90.00% specificity, and an MCC of 0.90.
• Five-factor solution score (5FSS) resulted in 90.00% accuracy, 96.00% sensitivity, 84.00% specificity, and an MCC of 0.81.
• Observed distinct differences in aliphatic and aromatic amino acid content between the two nitrilase groups.

Conclusions:

• Machine learning, particularly using PAAC, is effective for classifying nitrilases.
• The developed approach accurately predicts whether a nitrilase is aliphatic or aromatic based on its amino acid sequence.
• The computational scripts are available on GitHub for broader accessibility and application.