Towards a mechanism-aware multi-label prediction framework for antimicrobial peptides (DDLS PhD project)

Research Project, 2026 – 2030

Antimicrobial peptides (AMPs) are potent antimicrobial molecules that occur in all domains of life. Known for being part of the innate immune system, AMPs possess a plethora of activities including broad-spectrum antibacterial and antibiofilm activity as well as antifungal, antiviral, antiparasitic, and immunomodulatory properties. Their broad applicability and low propensity to generate microbial resistance have promoted AMPs as promising alternative to antibiotics and indeed, several AMPs have been successful in clinical applications. However, no other class of antimicrobial has been so thoroughly misunderstood as AMPs. For many decades, the mechanisms of action of AMPs have been simplified and broken down to unspecific destruction of the cell membrane, leading to a persistent image of these molecules as membrane pore formers. While this notion still permeates large parts of the scientific literature, it has been unambiguously shown that many AMPs do not form pores at all or only under artificial conditions that do not reflect infection. This revelation has prompted the realization that large parts of available data do not correctly capture AMP modes of action, limiting their usability for data-driven approaches to understanding their action. Our team has access to three unique datasets, composed of multi-omics, microscopy, and spectroscopy data, that have been meticulously designed and carefully curated and standardized to reflect the full complexity of AMP modes of action on microbial cells as accurately as possible. These curated datasets provide an optimal foundation to build the first mechanism-aware multi-label prediction framework for AMPs, aimed at providing deep insight into their biological modes of action and assisting future efforts to elucidate their mechanisms.