FrogPCSP: a propeptide cleavage site predictor for frog antimicrobial peptides.

Area of Science:

• Biochemistry
• Molecular Biology
• Bioinformatics

Background:

• Frogs secrete antimicrobial peptides (AMPs) for defense, stored as propeptides.
• Proprotein convertases (PCs) cleave propeptides into bioactive peptides and spacer peptides before secretion.
• Accurate identification of cleavage sites is crucial for predicting frog AMPs.

Purpose of the Study:

• To develop and validate Frog Propeptide Cleavage Site Predictor (FrogPCSP), an SVM-based tool for identifying propeptide cleavage sites in frog AMPs.
• To assess FrogPCSP's performance against existing prediction methods.
• To utilize FrogPCSP for discovering novel frog AMPs and characterizing their components.

Main Methods:

• Developed a Support Vector Machine (SVM) model (FrogPCSP) trained on known cleavage sites.
• Evaluated FrogPCSP using 10-fold cross-validation on 424 positive and 2488 negative cleavage sites.
• Compared FrogPCSP performance against PSSM and ProP predictors using AUC values.
• Applied FrogPCSP to analyze 595 unlabeled frog AMP sequences from UniProtKB/TrEMBL.

Main Results:

• FrogPCSP achieved high performance metrics: 0.981 global accuracy, 0.937 precision, 0.928 recall, and 0.933 F1-score.
• FrogPCSP demonstrated superior predictive performance (AUC=0.990) compared to PSSM (AUC=0.974) and ProP (AUC=0.905).
• Analysis of unlabeled sequences identified 926 putative cleavage sites, yielding distinct clusters of positively charged bioactive peptides and negatively charged spacer peptides.

Conclusions:

• FrogPCSP is a highly accurate and effective tool for predicting propeptide cleavage sites in frog AMPs.
• The predictor aids in the discovery of novel frog AMPs and enhances understanding of their structure-function relationships.
• Physicochemical profiling of predicted peptides supports their classification as bioactive or spacer peptides, validating the predictor's utility.