CS-AMPPred: An Updated SVM Model for Antimicrobial Activity Prediction in Cysteine-Stabilized Peptides

Abstract

The antimicrobial peptides (AMP) have been proposed as an alternative to control resistant pathogens. However, due to multifunctional properties of several AMP classes, until now there has been no way to perform efficient AMP identification, except through in vitro and in vivo tests. Nevertheless, an indication of activity can be provided by prediction methods. In order to contribute to the AMP prediction field, the CS-AMPPred (Cysteine-Stabilized Antimicrobial Peptides Predictor) is presented here, consisting of an updated version of the Support Vector Machine (SVM) model for antimicrobial activity prediction in cysteine-stabilized peptides. The CS-AMPPred is based on five sequence descriptors: indexes of (i) α-helix and (ii) loop formation; and averages of (iii) net charge, (iv) hydrophobicity and (v) flexibility. CS-AMPPred was based on 310 cysteine-stabilized AMPs and 310 sequences extracted from PDB. The polynomial kernel achieves the best accuracy on 5-fold cross validation (85.81%), while the radial and linear kernels achieve 84.19%. Testing in a blind data set, the polynomial and radial kernels achieve an accuracy of 90.00%, while the linear model achieves 89.33%. The three models reach higher accuracies than previously described methods. A standalone version of CS-AMPPred is available for download at <http://sourceforge.net/projects/csamppred/> and runs on any Linux machine.

Expression, purification and characterization of the recombinant cysteine-rich antimicrobial peptide snakin-1 in Pichia pastoris

“…Many AMPs are cysteine-stabilized and these have been classified into several families [14], including cyclotides [15], defensins [16], hevein-like peptides [17], lipid transfer proteins [18], and snakins [11,19]. The snakin family is a novel plant antimicrobial peptide family which shows good similarity with the members of the gibberellic acid stimulated transcript (GAST) and gibberellic acid stimulated in Arabidopsis (GASA) protein families in Arabidopsis [20].…”

Mapping membrane activity in undiscovered peptide sequence space using machine learning

“…Prior AMP machine-learning studies have focused primarily on empirical quantitative structure activity relationship (QSAR) models to evaluate large pools of candidate sequences and identify AMP candidates with improved minimum inhibitory concentrations (MICs) (19)(20)(21). QSAR models for AMP discovery use a variety of statistical learning approaches, including multiple linear regression, linear discriminant analysis, principal component analysis, partial least-squares regression, artificial neural networks (ANN), support vector classifiers (SVC)/support vector machines (SVM), quantitative matrices (QM), hidden Markov models (HMM), and random forests (RFs) (21)(22)(23)(24).…”

In Vivo Efficacy of Anuran Trypsin Inhibitory Peptides against Staphylococcal Skin Infection and the Impact of Peptide Cyclization

“…Variations in the antimicrobial susceptibility testing methods have been found to have significant effects on the MICs of compounds, as demonstrated using the cysteine-rich antimicrobial peptide protegrin-1, for example (52). Differences are also dependent on the characteristics of the antimicrobial agents, including hydrophobic moments, exposed charges, amphipathicity, and peptide flexibility (53,54). pYR was found to be the most active SLF peptide, while others had moderate-to-poor antimicrobial activity.…”