Antimicrobial peptide–drug conjugates: rational design, pharmacological evaluation, and translational perspectives

Antimicrobial peptide–drug conjugates (APDCs) combine peptide and antibiotic functions within a single covalent construct. APDCs share design features with antibody–drug conjugates, but the field lacks an integrated account of how such conjugates should be designed and evaluated. This review examines that process from scaffold selection to translational precedent. Design decisions are treated as medicinal chemistry: which antimicrobial peptide scaffold to start from and how to stabilize it, which payload to carry, whether the linker should be cleavable, conditionally stable, or non-cleavable, and whether engaging one or two targets yields a better antibacterial profile. Pharmacological evidence is then assessed for representative vancomycin, β-lactam, aminoglycoside, and fluoroquinolone conjugates, with experimentally demonstrated mechanisms distinguished from those that are indirectly inferred, computationally predicted, or hypothesized. In series reported through July 2026, conjugation produced large MIC reductions relative to inactive or resistant parent antibiotics, including in vancomycin-resistant enterococci, and in some cases restored activity against strains refractory to both parent molecules; comparisons with active standards of care were limited. By contrast, pharmacokinetic characterization is nearly absent, immunomodulatory and anti-biofilm mechanisms are only beginning to be dissected, in vivo efficacy has been reported for only seven APDC programs, and no dedicated toxicology study was identified. Two adjacent programs—the approved siderophore cephalosporin cefiderocol and the discontinued antibody–antibiotic conjugate DSTA4637S—serve as translational precedents, illustrating the manufacturing, safety, and regulatory hurdles an APDC must clear. Within these constraints, experimentally validated artificial-intelligence-assisted sequence design may support peptide modules for future APDCs, although our literature search did not identify an AI-designed APDC.