StackPVP: a stacked ensemble classification framework for predicting phage virion proteins using integrated...

Area of Science:

• Computational biology
• Bioinformatics
• Microbiology

Background:

• Phage therapy is a promising antibacterial strategy.
• Phage virion proteins (PVPs) are essential for host cell recognition and binding.
• Accurate identification of PVPs is critical for developing novel antibacterial agents.

Purpose of the Study:

• To introduce StackPVP, a computational approach for improved PVP identification.
• To integrate ensemble learning methods with evolutionary features for enhanced accuracy.
• To aid in phage genome annotation and advance phage therapy research.

Main Methods:

• Extracted features including AAC-PSSM, DPC-PSSM, Pse-PSSM, and PSSM-COM from PSSMs.
• Employed feature selection methods to identify an optimal feature subset.
• Utilized 12 base machine learning classifiers and selected random forest as the meta-classifier in a stacking model.

Main Results:

• StackPVP achieved an Area Under the Curve (AUC) of 94.26% on the test dataset.
• Demonstrated improved specificity compared to existing methods.
• The random forest meta-classifier in the stacking model showed superior performance.

Conclusions:

• StackPVP offers a valuable computational tool for accurate PVP identification.
• The approach can significantly assist in phage genome annotation.
• StackPVP contributes to the advancement of phage therapy and antibacterial research.