i2APP

i2APP identifies antiparasitic peptides (APPs) using a two-step machine learning framework to predict peptide antiparasitic activity.

Key Features:

• Two-step machine learning framework: Dual-phase strategy that balances positive and negative samples using random under-sampling and improves APP prediction accuracy.
• Feature extraction: Extracts physical/chemical features and terminus-based features from peptide sequences.
• Higher-level feature construction: Performs initial classification with Light Gradient Boosting Machine (LGBM) and Support Vector Machine (SVM) to generate 264-dimensional higher-level features.
• Feature selection: Uses Maximal Information Coefficient (MIC) to select the most informative features based on MIC values.
• Optimized classification: Applies a second-phase SVM classifier in the refined feature space for final APP prediction.
• Performance: On independent datasets reports an accuracy of 0.913 and an Area Under the Curve (AUC) of 0.935.

Scientific Applications:

• Parasitology research: Predicts and identifies peptides with antiparasitic properties to support studies of parasite biology and host–parasite interactions.
• Therapeutic peptide discovery: Prioritizes candidate peptides for development of new antiparasitic agents and potential treatments for parasitic infections.

Methodology:

Balancing the training dataset using random under-sampling; extracting physical/chemical and terminus-based multi-level features from peptide sequences; performing initial classification with LGBM and SVM to derive 264-dimensional higher-level features; selecting optimal features using Maximal Information Coefficient (MIC); and conducting a second classification phase with an SVM in the refined feature space.