University of Pennsylvania researchers disclosed ApexGO, an AI-guided method to optimize peptide antibiotics from a small starting set rather than screening massive candidate libraries. The approach iteratively proposes modifications, then steers optimization using a predictive algorithm. The team reported that most AI-generated peptides suppressed bacterial growth in lab tests, with a majority outperforming the parent peptides; in mice, two ApexGO-designed candidates reduced bacterial counts comparable to polymyxin B. The results were published in Nature Machine Intelligence. For biotech R&D, the dataset is a concrete example of AI-assisted “closed-loop” chemistry where predictions translate to biological function, potentially compressing time-to-lead selection for antimicrobial resistance programs.