Unsupervised learning elucidates the interplay between conformational flexibility and aggregation in synergistic antimicrobial peptides

Abstract

Synergy between antimicrobial peptides (AMPs) may be the key to their evolutionary success and could be exploited to develop more potent antibacterial agents. One of the factors thought to be essential for AMP potency is their conformational flexibility, but characterizing the diverse conformational states of AMPs experimentally remains challenging. Here we introduce a method for characterising the conformational flexibility of AMPs and provide new insights into how the interplay between conformation and aggregation in synergistic AMP combinations yields emergent properties. We use unsupervised learning and molecular dynamics simulations to show that mixing two AMPs from the Winter Flounder family (pleurocidin (WF2) & WF1a) constrains their conformational space, reducing the number of distinct conformations adopted by the peptides, most notably for WF2. The aggregation behaviour of the peptides is also altered, favouring the formation of higher-order aggregates upon mixing. Critically, the interaction between WF1a and WF2 influences the distribution of WF2 conformations within aggregates, revealing how WF1a can modulate WF2 behaviour. Our work paves the way for deeper understanding of the synergy between AMPs, a fundamental process in nature.