Vaxign-ML

Vaxign-ML predicts bacterial protective antigens (BPAgs) using supervised machine learning within a reverse vaccinology framework to prioritize vaccine candidates.

Key Features:

• Target: Focuses on predicting bacterial protective antigens (BPAgs) for vaccine candidate selection.
• Machine learning paradigm: Employs supervised machine learning for antigen prediction.
• Algorithms tested: Evaluated five different machine learning methods.
• Feature types: Uses biological and physiochemical features derived from well-defined training data.
• Validation strategies: Applies nested 5-fold cross-validation and leave-one-pathogen-out validation for performance assessment.
• Top-performing algorithm: eXtreme Gradient Boosting (XGBoost) emerged as the most effective algorithm.
• Benchmark comparisons: Compared predictive performance against Vaxign, VaxiJen, Antigenic, an SVM-based method, and an epitope-based approach using a high-quality benchmark dataset.
• Approach: Integrates reverse vaccinology principles with machine learning-based antigen prioritization.

Scientific Applications:

• Vaccine candidate prioritization: Prioritizes bacterial antigen candidates for vaccine design and experimental validation.
• Emerging pathogen prediction: Assesses model generalizability to new pathogens via leave-one-pathogen-out validation.
• Method benchmarking: Enables comparative evaluation of antigen prediction methods using a standardized benchmark dataset.

Methodology:

Five machine learning methods were tested using biological and physiochemical features from well-defined training data, with nested 5-fold cross-validation and leave-one-pathogen-out validation for performance evaluation; eXtreme Gradient Boosting achieved superior accuracy in comparisons against Vaxign, VaxiJen, Antigenic, an SVM-based method, and an epitope-based approach on a high-quality benchmark dataset.