Understanding and design of beta-sheet forming antimicrobial peptides

Antimicrobial resistance is a growing problem that must be combatted via the design of new antimicrobial agents. A difficult prospect, such design requires both searching rationally through uncharted areas in biomolecular search space and a heightened understanding of the molecular-level mechanisms by which different agents act to inhibit bacterial growth. An attractive line of research in both of these lines is antimicrobial peptides (AMPs), short amphiphilic peptides, many of which are produced by the innate immune systems of diverse organisms, and many of which are thought to act via membrane destabilization. While a great deal of attention has been paid to helix-forming AMPs, there are other important categories, including AMPs that aggregate into beta sheets on the surface of cell membranes and potentially indiscriminately strain the membrane, rather than forming pores. In this talk, I will report on our work in devising a workflow for the active learning design of beta-sheet-forming AMPs. I will discuss how we have employed emergent deep learning techniques in biomolecular de novo design to devise an initial estimate of a rational search space for biomolecular design and how in conjunction we have employed multi-scale molecular dynamics to study variants of a synthetic beta-sheet forming antimicrobial peptide at an atomistic level and at a coarse-grained level. We show charge-dependent self-interactions and how those interactions dictate the behavior in conjunction with a membrane. Overall, this work sheds light on methods for combatting antimicrobial resistance through a rational exploration of biomolecular search spaces.