Membrane-active peptides (MAPs) are a promising platform technology with a plethora of potential biomedical applications, from antimicrobial agents to drug delivery. The enormous combinatorial space of MAP sequences, however, has limited the rate of discovery in this area. The ROADMAP project at the National Institute of Standards and Technology (NIST) aims to overcome these limitations by leveraging the NIST Center for Neutron Research (NCNR) world-leading neutron reflectometry (NR) capabilities in combination with advances in AI-driven autonomous experimentation. In this presentation, I will introduce the ROADMAP project and present early progress. Initial topic models of antimicrobial peptide sequences discover motif families with potentially important roles in antimicrobial function; generative AI models produce candidate MAPs for subsequent measurements in a closed-loop fashion. On the measurement side, an automated liquid handling system for the high-speed CANDOR reflectometer at NCNR has established on-the-fly sample generation which significantly improves experimental throughput. Measurement results of antimicrobial activity and structural effects of the generated peptides on lipid membranes will be discussed. As a data-driven project at the NCNR national user facility, ROADMAP welcomes peptides to characterize from the community, and mechanisms for engagement will be presented.
Hoogerheide et al. (Sun,) studied this question.