Abstract

Protegrin antimicrobial peptides (AMP) possess a high activity against a variety of microorganisms. In the present contribution, we analyse the structural requirements of protegrin analogues reported by Ostberg and Kaznessis (Peptides 2005; 26: 197) for having antimicrobial activity against several microbial species by using interpretable QSAR models. Models were carried out using multiple linear regression (MLR) combined with genetic algorithm (GA) and smoothed amino acid sequence properties were employed for characterizing the peptide dataset. The main advantage of smoothing process is the alteration of local amino acid properties by the properties of the amino acids in the closer neighbourhood. We report models encompassing different characteristics for describing the activities against different microbial species. Our results suggest the existence of specific mechanisms of action for protegrin analogues against different microbial species.