T4SE-XGB

T4SE-XGB predicts type IV secreted effectors (T4SEs) from protein sequences using an eXtreme gradient boosting (XGBoost) classifier to support analysis of bacterial pathogenesis mediated by the type IV secretion system (T4SS).

Key Features:

• Algorithmic Approach: Uses XGBoost (eXtreme Gradient Boosting) as the primary classifier.
• Feature Optimization: Extracts 20 types of protein sequence features and selects an optimal subset using the ReliefF algorithm.
• Cross-Validation: Validates predictive performance using 5-fold cross-validation.
• Interpretability: Applies SHAP (SHapley Additive exPlanations) to quantify individual feature contributions to predictions.

Scientific Applications:

• T4SE Prediction: Identifies candidate proteins translocated by the T4SS to aid study of bacterial virulence mechanisms.
• Method Development: Serves as a framework for building machine-learning-based effector prediction methods in computational biology.

Methodology:

Extraction of 20 protein sequence feature types; feature refinement using ReliefF; model training with XGBoost; performance assessment via 5-fold cross-validation; interpretation of feature impacts using SHAP.