Prediction of Phage Virion Proteins Using Machine Learning Methods.

Area of Science:

• Microbiology
• Bioinformatics
• Machine Learning

Background:

• Antimicrobial resistance (AMR) poses a significant global health threat, necessitating urgent development of antibiotic alternatives.
• Bacteriophages (phages) and their derived products, such as phage virion proteins (PVPs), are emerging as promising therapeutic agents against AMR bacteria.
• Identifying novel PVPs is crucial for developing effective phage-based antibacterial strategies.

Purpose of the Study:

• To develop a machine learning-based method for predicting phage virion proteins (PVPs) from protein sequences.
• To evaluate the performance of various machine learning algorithms in identifying PVPs.
• To provide a user-friendly web server for accessible PVP prediction.

Main Methods:

• Utilized basic and ensemble machine learning techniques.
• Employed protein sequence composition features for model training.
• Trained and validated models using known phage protein sequences.

Main Results:

• The Gradient Boosting Classifier (GBC) achieved the highest accuracy, reaching 80% on the training dataset and 83% on an independent dataset.
• The developed method demonstrated superior performance compared to existing approaches on the independent dataset.
• A freely accessible web server was created for predicting PVPs from phage protein sequences.

Conclusions:

• Machine learning, specifically GBC, is effective for predicting PVPs, offering a valuable tool in the fight against AMR.
• The developed web server facilitates large-scale PVP prediction, supporting experimental research and the discovery of new antibacterial agents.
• PVPs represent a significant area for developing novel therapeutics to address the challenge of antimicrobial resistance.