Prediction of improved antimicrobial mastoparan derivatives by 3D-QSAR-CoMSIA/CoMFA and computational mutagenesis

Abstract

Antimicrobial peptides are an important class of therapeutic agents used against a wide range of pathogens such as gram-negative and -positive bacteria, fungi, and viruses. The minimal inhibitory concentration at the level of the pathogen membrane is a major determinant of the pharmacokinetic behavior and, consequently, it can affect their antimicrobial activity. Here we generated quantitative structure-activity relationship models (3D-QSAR—comparative molecular field analysis/comparative molecular similarity indices analysis) using a database of 33 mastoparan analogs, antimicrobial peptides with known experimental activity, and further used these models to predict the minimal inhibitory concentration for 18 new mastoparan analogs, obtained by computational mutagenesis. We discuss two options for structural alignment of mastoparan analogs: superposition of Cα trace atoms or superposition of all backbone atoms. Significant values of the cross-validated correlation q 2 (higher than 0.60) and the fitted correlation r 2 (higher than 0.90) of our models indicate that they are reliable enough for activity prediction in the case of new derivatives. This allows us to identify compounds with possibly enhanced antimicrobial activity against Bacillus subtilis, which are suggested for further experimental studies.