engleski

Positional specific amino acid attributes improve the computational design of antimicrobial peptides

A novel method and web server for prediction of TI values for helical antimicrobial peptides isolated from different frog species is designed based on amino acid frequencies and motifs for antibiotics and hemolytics. We find that alpha-amphipathic segments predicted by the Split server (http://split.pmfst.hr/split/) and G, Q, H, D, E amino acid frequencies are positively correlated, while high preferences for transmembrane helix conformation are negatively correlated with measured TI values. A composite index is developed as a specific combination of these positional and frequency attributes. This composite index is additionally combined with D-predictor described in [1], and very good correlation was obtained between experimental and calculated TI values on the training set containing 36 nonhomologous peptides (less than 70% pairwise identity). Correlation coefficient is ~ 0.9 both on linear and log-log scale. Finally, we performed prediction of TI values on 54 peptides from external set which resulted in correlation of about 0.85 on linear and log-log scale. [1] Juretić D, Vukičević D, Ilić, N, Antcheva N, Tossi A (2009) J Chem Inf Model 49, 2873

amino acid attributes; antimicrobial peptides; design

nije evidentirano