Krein support vector machine classification of antimicrobial peptides.

Area of Science:

• Computational biology
• Machine learning
• Bioinformatics

Background:

• Antimicrobial resistance (AMR) is a growing global health threat.
• Experimental identification of antimicrobial peptides (AMPs) is resource-intensive.
• Computational prediction of AMPs can accelerate drug discovery.

Purpose of the Study:

• To develop and evaluate Kreĭn-SVM models for predicting antimicrobial peptide activity.
• To utilize sequence similarity functions like Levenshtein distance and local alignment score.
• To enable rapid in silico screening of candidate AMPs.

Main Methods:

• Implementation of Kreĭn-SVM models for AMP classification.
• Training models on large peptide datasets for general antimicrobial activity prediction.
• Curating microbe-specific datasets for targeted activity prediction.

Main Results:

• Achieved high AUC values (0.967, 0.863) for general AMP prediction, outperforming baselines.
• Demonstrated strong performance for microbe-specific AMP prediction (AUC 0.982, 0.891).
• Developed web applications for predicting general and microbe-specific AMP activities.

Conclusions:

• Kreĭn-SVM models offer a powerful and efficient approach for AMP discovery.
• The methodology successfully predicts both general and microbe-specific antimicrobial activities.
• Accessible web tools facilitate the application of these predictive models.