Development of tissue-specific homing peptides using probabilistic machine learning

Abstract

A central challenge for curing complex diseases is inefficient drug delivery. One potential solution is to use homing peptides that target unique tissue markers for selective drug delivery, increasing therapeutic effectiveness and decreasing side effects. This thesis presents an integrated in silico and in vitro workflow that accelerates the discovery and optimization of receptor-specific homing peptides from months and years to weeks. High-copy T7-phage display proved the importance of multivalency, showing higher binding in all cases. These results show that generative modelling can learn receptor-specific pharmacophores and potentially yield functional ligands, offering a hand to time-consuming receptor-aware peptide design.