Co-AMPpred

Co-AMPpred predicts antimicrobial peptides (AMPs) from amino-acid sequences using a composition-based model that integrates sequence and physicochemical features to support discovery of novel antimicrobial agents.

Key Features:

• Composition-based model: Integrates sequence and physicochemical features of amino-acid residues for peptide representation.
• Sequence and physicochemical descriptors: Uses compositional and residue-level physicochemical features as input for prediction.
• Machine-learning classification: Employs machine-learning algorithms to classify sequences as AMPs or non-AMPs.
• Boruta feature selection: Applies the Boruta algorithm to identify discriminative biological features for the model.
• Benchmark datasets: Trained and validated on benchmark datasets compiled from previous studies.
• Cross-validation: Validated using stratified tenfold cross-validation on benchmark datasets.
• Independent evaluation: Assessed on an independent holdout test dataset.
• Performance metrics: Reported accuracy of 80.8% and area under the ROC curve (AUC) of 0.871 on the independent test set.
• Biological context: Targets antimicrobial peptides (AMPs), which are oligopeptides involved in innate immunity and regulation of host processes such as wound healing and apoptosis.

Scientific Applications:

• AMP discovery: Prioritizes candidate antimicrobial peptides for experimental validation and therapeutic development.
• Antibiotic-resistance research: Supports identification of novel AMPs as alternative strategies to conventional antibiotics.
• Bioinformatics and microbiology studies: Provides computational predictions to aid mechanistic and comparative analyses of AMP sequences.

Methodology:

Co-AMPpred uses a composition-based model integrating sequence and physicochemical amino-acid features, applies the Boruta feature-selection algorithm and machine-learning classifiers, and was validated by stratified tenfold cross-validation and independent holdout testing on benchmark datasets.