PhotoRaPID: Direct selection of light-responsive macrocyclic peptides from trillion-member libraries

Abstract

Photopharmacology seeks to enable precise optical control of protein function, but the discovery of light-responsive ligands remains challenging, particularly for biologically and therapeutically relevant targets lacking pre-existing binders. Current strategies rely on derivatisation of known ligands with photoswitches, significantly limiting target scope. As a result, a generalisable approach for the discovery of photoswitchable ligands is needed by the field. Here we develop PhotoRaPID, a de novo mRNA display platform that uses genetic-code reprogramming to install azopyrazole photoswitches into peptide backbones, enabling selection of photoswitchable macrocyclic peptides from libraries exceeding 1012 members. Photoselection enriches longlived Z-state binders, yielding macrocycles with low-nanomolar affinities, excellent Z/E activity differentials, and potent target inhibition. We validate the platform on the cancer-relevant and functionally diverse proteins TIGIT and PADI2, for which no photoswitchable or macrocyclic peptide ligands have been reported, and demonstrate light-gated target engagement and function. A newly developed unsupervised clustering framework for sequencing data resolves sequence convergence and family-level motifs, enabling prioritisation of diverse binders and shifting post-selection analysis from qualitative to quantitative assessment. Together, these results establish a generalisable strategy for discovering light-responsive macrocyclic peptides. More broadly, PhotoRaPID makes photochemical state dependence a selectable molecular phenotype, enabling direct discovery of light-responsive ligands across diverse target classes rather than retrospective photoswitch installation into known binders.