antimpmod

antimpmod predicts the antimicrobial potential of chemically modified peptides (ModAMPs) to support design and evaluation of therapeutic peptides.

Key Features:

• Dataset and Training: The model was trained and evaluated on a dataset of 948 antimicrobial and 931 non-antimicrobial peptides containing chemically modified and natural residues.
• Structural Prediction: Tertiary peptide structures were predicted using PEPstrMOD.
• Feature Computation: A wide array of chemical and physical structural features were computed from predicted structures using PaDEL.
• Predictive Modeling: A support vector machine (SVM) model was developed, achieving a maximum Matthews Correlation Coefficient (MCC) of 0.84 and accuracy of 91.62% on training data, and MCC of 0.80 and accuracy of 89.89% on validation data.

Scientific Applications:

• Antimicrobial Peptide Design: Supports design and optimization of novel antimicrobial peptides by predicting effects of chemical modifications.
• Resistance Mitigation: Predicts modified-peptide candidates that may overcome antibiotic-resistant strains.
• Drug Discovery: Identifies promising peptide candidates early in the therapeutic development pipeline.

Methodology:

PEPstrMOD was used for tertiary-structure prediction, PaDEL for computation of chemical and physical structural features, and an SVM model was trained and evaluated on the 948 antimicrobial / 931 non-antimicrobial peptide dataset.