Predicting the Minimal Inhibitory Concentration for Antimicrobial Peptides with Rana-Box Domain

Abstract

The global spreading of multidrug resistance has motivated the search for new antibiotic classes including different types of antimicrobial peptides (AMPs). Computational methods for predicting activity in terms of the minimal inhibitory concentration (MIC) of AMPs can facilitate "in silico" design and reduce the cost of synthesis and testing. We have used an original method for separating training and test data sets, both of which contain the sequences and measured MIC values of non-homologous anuran peptides having the Rana-box disulfide motif at their C-terminus. Using a more flexible profiling methodology (sideways asymmetry moment, SAM) than the standard hydrophobic moment, we have developed a two-descriptor model to predict the bacteriostatic activity of Rana-box peptides against Gram-negative bacteria--the first multilinear quantitative structure-activity relationship model capable of predicting MIC values for AMPs of widely different lengths and low identity using such a small number of descriptors. Maximal values for SAMs, as defined and calculated in our method, furthermore offer new structural insight into how different segments of a peptide contribute to its bacteriostatic activity, and this work lays the foundations for the design of active artificial AMPs with this type of disulfide bridge.