ProtEC: A Transformer Based Deep Learning System for Accurate Annotation of Enzyme Commission Numbers.

Area of Science:

• Bioinformatics
• Computational Biology
• Enzymology

Background:

• Next-generation sequencing has created vast protein databases.
• Manual annotation of protein sequences is time-consuming and labor-intensive.
• Automated annotation algorithms, including deep learning, are crucial for biological data interpretation.

Purpose of the Study:

• To develop a novel transformer-based deep learning model for predicting Enzyme Commission (EC) numbers.
• To achieve state-of-the-art accuracy in enzyme annotation from protein sequences.
• To assess the model's robustness across diverse sequence datasets and varying training sizes.

Main Methods:

• A transformer-based deep learning architecture was employed.
• The model was trained to predict Enzyme Commission numbers directly from full-scale protein sequences.
• Performance was evaluated on clustered split datasets with structurally dissimilar distributions.

Main Results:

• The proposed model achieved state-of-the-art accuracy in predicting EC numbers.
• The model demonstrated strong performance on structurally dissimilar datasets, indicating deep pattern recognition.
• Accuracy was maintained with reduced training data and was independent of sequence length.

Conclusions:

• The transformer-based model offers a highly accurate and robust solution for automated enzyme annotation.
• Its ability to generalize across sequence variations makes it suitable for diverse biomedical applications.
• This approach significantly advances the utility of large-scale protein sequence databases.