Project title: Learning by Testing: A Lab-in-the-Loop Strategy for Antimicrobial Peptide Discovery

Brief project description: The vast diversity of antimicrobial peptide sequences makes exhaustive screening impractical. For example, a 20-amino-acid sequence space contains roughly 3.5 × 10²⁹ possibilities. Machine learning can help identify promising candidates, but its success depends on large, reliable datasets. Public data are often limited and inconsistent because experiments use different conditions. We are therefore developing a lab-in-the-loop approach that pairs deep learning with high-throughput experiments under consistent conditions. Experimental results train the model, which then designs the next peptide library for testing. Repeating this cycle guides exploration toward more promising regions of sequence space, with the goal of identifying increasingly enriched libraries of active peptides.