StackDPPred: Multiclass prediction of defensin peptides using stacked ensemble learning with optimized features.

Area of Science:

• Biochemistry and Molecular Biology
• Computational Biology
• Drug Discovery

Background:

• Antimicrobial peptides (AMPs), including defensins, are crucial for host defense against pathogens.
• Traditional methods for identifying defensin peptides (DPs) are laborious and costly.
• Computational approaches offer a more efficient alternative for DPs prediction.

Purpose of the Study:

• To develop a novel ensemble-based computational model, StackDPPred, for predicting defensin peptide properties.
• To enhance the accuracy and efficiency of identifying functional DPs and their families.
• To accelerate the screening process for peptide-based drug discovery.

Main Methods:

• Peptide sequences encoded using Split Amino Acid Composition (SAAC), Segmented Position Specific Scoring Matrix (SegPSSM), Histogram of Oriented Gradients-based PSSM (HOGPSSM), and FEGS descriptors.
• Principal Component Analysis (PCA) for feature selection.
• Stacking-based ensemble classifiers integrating machine learning algorithms.

Main Results:

• StackDPPred achieved significant improvements in prediction accuracy compared to existing methods (iDPF-PseRAAC and iDEF-PseRAAC).
• Ablation studies confirmed the robustness and efficacy of the stacking ensemble approach.
• Local Interpretable Model-agnostic Explanations (LIME) provided insights into feature contributions.

Conclusions:

• StackDPPred offers a powerful and accurate computational tool for DPs identification.
• The model can significantly accelerate the discovery of novel peptide-based therapeutics.
• This work contributes to advancing antimicrobial peptide research and drug development.